diff --git "a/openvino_model.xml" "b/openvino_model.xml"
--- "a/openvino_model.xml"
+++ "b/openvino_model.xml"
@@ -36,25 +36,13 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4" name="Constant_337416_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
+		<layer id="4" name="Constant_52539" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="5" name="Constant_337416" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
+				<port id="0" precision="FP32" />
 			</output>
 		</layer>
-		<layer id="6" name="ShapeOf_337402" type="ShapeOf" version="opset3">
+		<layer id="5" name="ShapeOf_52525" type="ShapeOf" version="opset3">
 			<data output_type="i64" />
 			<input>
 				<port id="0" precision="I64">
@@ -68,21 +56,21 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7" name="Constant_337404" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
+		<layer id="6" name="Constant_52527" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="8" name="Constant_337406" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
+		<layer id="7" name="Constant_52529" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="9" name="Gather_337407" type="Gather" version="opset8">
+		<layer id="8" name="Gather_52530" type="Gather" version="opset8">
 			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="I64">
@@ -99,31 +87,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="10" name="Constant_337409" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
+		<layer id="9" name="Constant_52532" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="11" name="Constant_337411" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
+		<layer id="10" name="Constant_52534" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="12" name="Constant_337413" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
+		<layer id="11" name="Constant_52536" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="13" name="Concat_337414" type="Concat" version="opset1">
+		<layer id="12" name="Concat_52537" type="Concat" version="opset1">
 			<data axis="0" />
 			<input>
 				<port id="0" precision="I64">
@@ -145,7 +133,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="14" name="Broadcast_337417" type="Broadcast" version="opset3">
+		<layer id="13" name="Broadcast_52540" type="Broadcast" version="opset3">
 			<data mode="numpy" />
 			<input>
 				<port id="0" precision="FP32" />
@@ -162,7 +150,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="15" name="ReadValue_330551" type="ReadValue" version="opset6">
+		<layer id="14" name="ReadValue_50309" type="ReadValue" version="opset6">
 			<data variable_id="past_key_values.9.valuepresent.9.value" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
 				<port id="0" precision="FP32">
@@ -181,13 +169,13 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="16" name="Constant_323442" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
+		<layer id="15" name="Constant_47835" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="17" name="Gather_323443" type="Gather" version="opset8">
+		<layer id="16" name="Gather_47836" type="Gather" version="opset8">
 			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="FP32">
@@ -210,8 +198,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="18" name="self.model.gpt_neox.embed_in.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="50280, 2560" offset="26" size="128716800" />
+		<layer id="17" name="self.model.gpt_neox.embed_in.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="50280, 2560" offset="28" size="128716800" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>50280</dim>
@@ -219,7 +207,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="19" name="Convert_21414" type="Convert" version="opset1">
+		<layer id="18" name="Convert_74164" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -234,8 +222,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="20" name="self.model.gpt_neox.embed_in.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="50280, 1" offset="128716826" size="50280" />
+		<layer id="19" name="self.model.gpt_neox.embed_in.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="50280, 1" offset="128716828" size="50280" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>50280</dim>
@@ -243,7 +231,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="21" name="Convert_21417" type="Convert" version="opset1">
+		<layer id="20" name="Convert_74167" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -258,7 +246,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="22" name="self.model.gpt_neox.embed_in.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="21" name="self.model.gpt_neox.embed_in.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -277,8 +265,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="23" name="self.model.gpt_neox.embed_in.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="50280, 1" offset="128767106" size="100560" />
+		<layer id="22" name="self.model.gpt_neox.embed_in.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="50280, 1" offset="128767108" size="100560" />
 			<output>
 				<port id="0" precision="FP16">
 					<dim>50280</dim>
@@ -286,7 +274,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="24" name="self.model.gpt_neox.embed_in.weight_compressed/fq_weights_0" type="Multiply" version="opset1">
+		<layer id="23" name="self.model.gpt_neox.embed_in.weight/fq_weights_0" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -305,7 +293,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="25" name="self.model.gpt_neox.embed_in.weight" type="Convert" version="opset1">
+		<layer id="24" name="__module.model.gpt_neox.embed_in/ov_ext::embedding/Convert" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
@@ -323,7 +311,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="26" name="__module.model.gpt_neox.embed_in/aten::embedding/Convert" type="Convert" version="opset1">
+		<layer id="25" name="__module.model.gpt_neox.embed_in/ov_ext::embedding/Convert_1" type="Convert" version="opset1">
 			<data destination_type="i32" />
 			<input>
 				<port id="0" precision="I64">
@@ -338,13 +326,13 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="27" name="__module.model.gpt_neox.embed_in/aten::embedding/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
+		<layer id="26" name="__module.model.gpt_neox.embed_in/ov_ext::embedding/Constant" type="Const" version="opset1">
+			<data element_type="i32" shape="" offset="0" size="4" />
 			<output>
 				<port id="0" precision="I32" />
 			</output>
 		</layer>
-		<layer id="28" name="__module.model.gpt_neox.embed_in/aten::embedding/Gather" type="Gather" version="opset8">
+		<layer id="27" name="__module.model.gpt_neox.embed_in/ov_ext::embedding/Gather" type="Gather" version="opset8">
 			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="FP32">
@@ -358,22 +346,22 @@
 				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="307,input.1">
+				<port id="3" precision="FP32" names="297,inputs_embeds">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="29" name="__module.model.gpt_neox.layers.0.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867670" size="4" />
+		<layer id="28" name="__module.model.gpt_neox.layers.0.input_layernorm/aten::layer_norm/Multiply" type="Const" version="opset1">
+			<data element_type="i32" shape="1" offset="128867668" size="4" />
 			<output>
 				<port id="0" precision="I32">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="30" name="__module.model.gpt_neox.layers.0.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
+		<layer id="29" name="__module.model.gpt_neox.layers.0.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
 			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
 			<input>
 				<port id="0" precision="FP32">
@@ -393,37 +381,17 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="31" name="Constant_299066_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="128867674" size="5120" />
+		<layer id="30" name="Constant_37287" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="128867672" size="10240" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="32" name="Constant_299066" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="33" name="__module.model.gpt_neox.layers.0.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
+		<layer id="31" name="__module.model.gpt_neox.layers.0.post_attention_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -445,37 +413,17 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="34" name="Constant_299067_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="128872794" size="5120" />
+		<layer id="32" name="Constant_37288" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="128877912" size="10240" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="35" name="Constant_299067" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="36" name="__module.model.gpt_neox.layers.0.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+		<layer id="33" name="__module.model.gpt_neox.layers.0.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -490,15 +438,15 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="441">
+				<port id="2" precision="FP32" names="473">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="37" name="self.model.gpt_neox.layers.0.mlp.dense_h_to_4h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 2560" offset="128877914" size="26214400" />
+		<layer id="34" name="self.model.gpt_neox.layers.0.mlp.dense_h_to_4h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 2560" offset="128888152" size="26214400" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>10240</dim>
@@ -506,7 +454,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="38" name="Convert_21444" type="Convert" version="opset1">
+		<layer id="35" name="Convert_74184" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -521,8 +469,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="39" name="self.model.gpt_neox.layers.0.mlp.dense_h_to_4h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 1" offset="155092314" size="10240" />
+		<layer id="36" name="self.model.gpt_neox.layers.0.mlp.dense_h_to_4h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 1" offset="155102552" size="10240" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>10240</dim>
@@ -530,7 +478,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="40" name="Convert_21447" type="Convert" version="opset1">
+		<layer id="37" name="Convert_74187" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -545,7 +493,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="41" name="self.model.gpt_neox.layers.0.mlp.dense_h_to_4h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="38" name="self.model.gpt_neox.layers.0.mlp.dense_h_to_4h.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -564,8 +512,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="42" name="self.model.gpt_neox.layers.0.mlp.dense_h_to_4h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="10240, 1" offset="155102554" size="20480" />
+		<layer id="39" name="self.model.gpt_neox.layers.0.mlp.dense_h_to_4h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="10240, 1" offset="155112792" size="20480" />
 			<output>
 				<port id="0" precision="FP16">
 					<dim>10240</dim>
@@ -573,7 +521,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="43" name="self.model.gpt_neox.layers.0.mlp.dense_h_to_4h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="40" name="self.model.gpt_neox.layers.0.mlp.dense_h_to_4h.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -592,7 +540,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="44" name="self.model.gpt_neox.layers.0.mlp.dense_h_to_4h.weight" type="Convert" version="opset1">
+		<layer id="41" name="__module.model.gpt_neox.layers.0.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
@@ -610,7 +558,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="45" name="__module.model.gpt_neox.layers.0.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="42" name="__module.model.gpt_neox.layers.0.mlp.dense_h_to_4h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -631,37 +579,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="46" name="Constant_299068_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 10240" offset="155123034" size="20480" />
+		<layer id="43" name="self.model.gpt_neox.layers.0.mlp.dense_h_to_4h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="10240" offset="155133272" size="20480" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.0.mlp.dense_h_to_4h.bias">
 					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="47" name="Constant_299068" type="Convert" version="opset1">
+		<layer id="44" name="__module.model.gpt_neox.layers.0.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
 			</rt_info>
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16">
 					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="48" name="__module.model.gpt_neox.layers.0.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1">
+		<layer id="45" name="__module.model.gpt_neox.layers.0.mlp.dense_h_to_4h/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -670,20 +612,18 @@
 					<dim>10240</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="446">
+				<port id="2" precision="FP32" names="478">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="49" name="__module.model.gpt_neox.layers.0.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
+		<layer id="46" name="__module.model.gpt_neox.layers.0.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
 			<data approximation_mode="ERF" />
 			<input>
 				<port id="0" precision="FP32">
@@ -693,15 +633,15 @@
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="447">
+				<port id="1" precision="FP32" names="479">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="50" name="self.model.gpt_neox.layers.0.mlp.dense_4h_to_h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 10240" offset="155143514" size="26214400" />
+		<layer id="47" name="self.model.gpt_neox.layers.0.mlp.dense_4h_to_h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 10240" offset="155153752" size="26214400" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>2560</dim>
@@ -709,7 +649,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="51" name="Convert_21454" type="Convert" version="opset1">
+		<layer id="48" name="Convert_74194" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -724,8 +664,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="52" name="self.model.gpt_neox.layers.0.mlp.dense_4h_to_h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="181357914" size="2560" />
+		<layer id="49" name="self.model.gpt_neox.layers.0.mlp.dense_4h_to_h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="181368152" size="2560" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>2560</dim>
@@ -733,7 +673,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="53" name="Convert_21457" type="Convert" version="opset1">
+		<layer id="50" name="Convert_74197" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -748,7 +688,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="54" name="self.model.gpt_neox.layers.0.mlp.dense_4h_to_h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="51" name="self.model.gpt_neox.layers.0.mlp.dense_4h_to_h.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -767,8 +707,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="55" name="self.model.gpt_neox.layers.0.mlp.dense_4h_to_h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="181360474" size="5120" />
+		<layer id="52" name="self.model.gpt_neox.layers.0.mlp.dense_4h_to_h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="181370712" size="5120" />
 			<output>
 				<port id="0" precision="FP16">
 					<dim>2560</dim>
@@ -776,7 +716,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="56" name="self.model.gpt_neox.layers.0.mlp.dense_4h_to_h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="53" name="self.model.gpt_neox.layers.0.mlp.dense_4h_to_h.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -795,7 +735,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="57" name="self.model.gpt_neox.layers.0.mlp.dense_4h_to_h.weight" type="Convert" version="opset1">
+		<layer id="54" name="__module.model.gpt_neox.layers.0.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
@@ -813,7 +753,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="58" name="__module.model.gpt_neox.layers.0.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="55" name="__module.model.gpt_neox.layers.0.mlp.dense_4h_to_h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -834,37 +774,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="59" name="Constant_299069_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="181365594" size="5120" />
+		<layer id="56" name="self.model.gpt_neox.layers.0.mlp.dense_4h_to_h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="2560" offset="181375832" size="5120" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.0.mlp.dense_4h_to_h.bias">
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="60" name="Constant_299069" type="Convert" version="opset1">
+		<layer id="57" name="__module.model.gpt_neox.layers.0.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
 			</rt_info>
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16">
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="61" name="__module.model.gpt_neox.layers.0.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1">
+		<layer id="58" name="__module.model.gpt_neox.layers.0.mlp.dense_4h_to_h/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -873,50 +807,28 @@
 					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="450,input.5">
+				<port id="2" precision="FP32" names="482,input.3">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="62" name="Constant_299070_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="181370714" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="63" name="Constant_299070" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="59" name="Constant_37289" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="181380952" size="10240" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="64" name="__module.model.gpt_neox.layers.0.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
+		<layer id="60" name="__module.model.gpt_neox.layers.0.input_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -938,37 +850,17 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="65" name="Constant_299071_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="181375834" size="5120" />
+		<layer id="61" name="Constant_37290" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="181391192" size="10240" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="66" name="Constant_299071" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="67" name="__module.model.gpt_neox.layers.0.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+		<layer id="62" name="__module.model.gpt_neox.layers.0.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -983,15 +875,15 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="316">
+				<port id="2" precision="FP32" names="382,hidden_states.1">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="68" name="self.model.gpt_neox.layers.0.attention.query_key_value.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 2560" offset="181380954" size="19660800" />
+		<layer id="63" name="self.model.gpt_neox.layers.0.attention.query_key_value.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 2560" offset="181401432" size="19660800" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>7680</dim>
@@ -999,7 +891,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="69" name="Convert_21424" type="Convert" version="opset1">
+		<layer id="64" name="Convert_74824" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -1014,8 +906,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="70" name="self.model.gpt_neox.layers.0.attention.query_key_value.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 1" offset="201041754" size="7680" />
+		<layer id="65" name="self.model.gpt_neox.layers.0.attention.query_key_value.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 1" offset="201062232" size="7680" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>7680</dim>
@@ -1023,7 +915,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="71" name="Convert_21427" type="Convert" version="opset1">
+		<layer id="66" name="Convert_74827" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -1038,7 +930,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="72" name="self.model.gpt_neox.layers.0.attention.query_key_value.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="67" name="self.model.gpt_neox.layers.0.attention.query_key_value.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -1057,8 +949,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="73" name="self.model.gpt_neox.layers.0.attention.query_key_value.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="7680, 1" offset="201049434" size="15360" />
+		<layer id="68" name="self.model.gpt_neox.layers.0.attention.query_key_value.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="7680, 1" offset="201069912" size="15360" />
 			<output>
 				<port id="0" precision="FP16">
 					<dim>7680</dim>
@@ -1066,7 +958,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="74" name="self.model.gpt_neox.layers.0.attention.query_key_value.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="69" name="self.model.gpt_neox.layers.0.attention.query_key_value.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -1085,11 +977,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="75" name="self.model.gpt_neox.layers.0.attention.query_key_value.weight" type="Convert" version="opset1">
+		<layer id="70" name="self.model.gpt_neox.layers.0.attention.query_key_value.weight/fq_weights_1/convert" type="Convert" version="opset1">
 			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
 			<input>
 				<port id="0" precision="FP16">
 					<dim>7680</dim>
@@ -1103,7 +992,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="76" name="__module.model.gpt_neox.layers.0.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="71" name="__module.model.gpt_neox.layers.0.attention.query_key_value/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -1124,37 +1013,17 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="77" name="Constant_299072_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 7680" offset="201064794" size="15360" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="78" name="Constant_299072" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</input>
+		<layer id="72" name="Constant_37291" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 7680" offset="201085272" size="30720" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="79" name="__module.model.gpt_neox.layers.0.attention.query_key_value/aten::linear/Add" type="Add" version="opset1">
+		<layer id="73" name="__module.model.gpt_neox.layers.0.attention.query_key_value/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -1169,25 +1038,22 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="323,qkv.1">
+				<port id="2" precision="FP32" names="389,qkv.1">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="80" name="__module.model.gpt_neox.layers.0.attention/prim::ListConstruct/Concat" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080154" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="74" name="__module.model.gpt_neox.layers.0.attention/prim::ListConstruct" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201115992" size="32" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="81" name="__module.model.gpt_neox.layers.0.attention/aten::view/Reshape" type="Reshape" version="opset1">
+		<layer id="75" name="__module.model.gpt_neox.layers.0.attention/aten::view/Reshape" type="Reshape" version="opset1">
 			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -1200,7 +1066,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="327,qkv.3">
+				<port id="2" precision="FP32" names="393,qkv.3">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -1208,230 +1074,39 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="82" name="Constant_269016" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="83" name="Constant_269019" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="84" name="Constant_269022" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="85" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="328">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="86" name="Constant_223" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="329">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="87" name="__module.model.gpt_neox.layers.0.attention/aten::permute/Transpose" type="Transpose" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="330,query.1">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="88" name="Constant_269028" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="89" name="Constant_269031" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="90" name="Constant_269034" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="91" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="337,q.1">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="92" name="self.model.gpt_neox.layers.0.attention.rotary_emb.cos_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201080346" size="81920" />
-			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.0.attention.rotary_emb.cos_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="93" name="self.model.gpt_neox.layers.0.attention.rotary_emb.cos_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="94" name="__module.model.gpt_neox.layers.0.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="76" name="Constant_710" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="95" name="Constant_269040" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="77" name="Constant_712" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="96" name="Constant_269043" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="78" name="Constant_714" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="97" name="Constant_269046" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="79" name="__module.model.gpt_neox/aten::slice/Reshape_1" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116032" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="98" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="80" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Slice" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -1440,17 +1115,20 @@
 					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="331">
+				<port id="5" precision="FP32" names="394">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -1458,15 +1136,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="99" name="Constant_237" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="81" name="Constant_723" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64" names="332">
+				<port id="0" precision="I64" names="395">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="100" name="__module.model.gpt_neox.layers.0.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
+		<layer id="82" name="__module.model.gpt_neox.layers.0.attention/aten::permute/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -1479,7 +1157,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="333,key.1">
+				<port id="2" precision="FP32" names="396,query.1">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -1487,8 +1165,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="101" name="__module.model.gpt_neox.layers.0.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="83" name="Constant_764" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="84" name="Constant_766" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="85" name="Constant_768" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="86" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Slice_1" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -1496,87 +1197,100 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-			</input>
-			<output>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="102" name="206" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="201162298" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64" names="206" />
+				<port id="5" precision="FP32" names="403,q.1">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="103" name="__module.model.gpt_neox.layers.0.attention/aten::size/Constant_2" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="87" name="__module.model.gpt_neox.rotary_emb/aten::to/Convert" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 10, 1" offset="201116080" size="40" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="FP32" names="352">
+					<dim>1</dim>
+					<dim>10</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="104" name="__module.model.gpt_neox.layers.0.attention/aten::size/Gather_2" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="88" name="ShapeOf_38678" type="ShapeOf" version="opset3">
+			<data output_type="i64" />
 			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="341" />
+				<port id="1" precision="I64">
+					<dim>2</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="105" name="Constant_337429_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
+		<layer id="89" name="Constant_38679" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="106" name="Constant_337429" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
+		<layer id="90" name="Constant_38680" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="1" precision="FP32" />
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="107" name="Constant_337422" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
-			<output>
+		<layer id="91" name="Gather_38681" type="Gather" version="opset8">
+			<data batch_dims="0" />
+			<input>
 				<port id="0" precision="I64">
+					<dim>2</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64" />
+			</input>
+			<output>
+				<port id="3" precision="I64" names="353">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="108" name="Constant_337424" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
+		<layer id="92" name="Constant_32655" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="109" name="Constant_337426" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
+		<layer id="93" name="Constant_32657" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="110" name="Concat_337427" type="Concat" version="opset1">
+		<layer id="94" name="__module.model.gpt_neox.rotary_emb/prim::ListConstruct" type="Concat" version="opset1">
 			<data axis="0" />
 			<input>
 				<port id="0" precision="I64">
@@ -1588,233 +1302,160 @@
 				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="3" precision="I64">
+					<dim>3</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="111" name="Broadcast_337430" type="Broadcast" version="opset3">
-			<data mode="numpy" />
+		<layer id="95" name="__module.model.gpt_neox.rotary_emb/aten::expand/Broadcast" type="Broadcast" version="opset3">
+			<data mode="bidirectional" />
 			<input>
-				<port id="0" precision="FP32" />
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>10</dim>
+					<dim>1</dim>
+				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>3</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="355,360,inv_freq_expanded,inv_freq_expanded.1">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
+					<dim>10</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="112" name="ReadValue_330425" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.0.keypresent.0.key" variable_type="f32" variable_shape="?,32,?,80" />
+		<layer id="96" name="204" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64" names="204" />
+			</output>
+		</layer>
+		<layer id="97" name="__module.model.gpt_neox.rotary_emb/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1">
 			<input>
-				<port id="0" precision="FP32">
+				<port id="0" precision="I64">
+					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
 				</port>
+				<port id="1" precision="I64" />
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="past_key_values.0.key">
+				<port id="2" precision="I64" names="357,358">
 					<dim>-1</dim>
-					<dim>32</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="113" name="Constant_323385" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64" />
-			</output>
-		</layer>
-		<layer id="114" name="Gather_323386" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="98" name="__module.model.gpt_neox.rotary_emb/aten::to/Convert_2" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
+				<port id="0" precision="I64">
 					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I32">
+					<dim>1</dim>
 					<dim>-1</dim>
 				</port>
-				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="3" precision="FP32">
+				<port id="1" precision="FP32" names="359,361,position_ids_expanded,position_ids_expanded.1">
 					<dim>-1</dim>
-					<dim>32</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="115" name="__module.model.gpt_neox.layers.0.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="99" name="__module.model.gpt_neox.rotary_emb/aten::matmul/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="false" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>10</dim>
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="116" name="__module.model.gpt_neox.layers.0.attention/aten::size/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="117" name="__module.model.gpt_neox.layers.0.attention/aten::size/Gather_3" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="343" />
-			</output>
-		</layer>
-		<layer id="118" name="__module.model.gpt_neox.layers.0.attention/aten::add_/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" names="342,346,347,seq_len.1" />
-			</output>
-		</layer>
-		<layer id="119" name="__module.model.gpt_neox.layers.0.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="120" name="__module.model.gpt_neox.layers.0.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
 			</input>
 			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="362">
+					<dim>-1</dim>
+					<dim>10</dim>
+					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="121" name="__module.model.gpt_neox.layers.0.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="100" name="__module.model.gpt_neox.rotary_emb/aten::transpose/Constant" type="Const" version="opset1">
+			<data element_type="i32" shape="3" offset="201116120" size="12" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="I32">
+					<dim>3</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="122" name="__module.model.gpt_neox.layers.0.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="101" name="__module.model.gpt_neox.rotary_emb/aten::transpose/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+					<dim>-1</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
+				<port id="1" precision="I32">
+					<dim>3</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="350,353,cos.1">
+				<port id="2" precision="FP32" names="363">
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="123" name="__module.model.gpt_neox.layers.0.attention/aten::index/Convert" type="Convert" version="opset1">
-			<data destination_type="i32" />
+		<layer id="102" name="__module.model.gpt_neox.rotary_emb/aten::cat/Concat" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
-				<port id="0" precision="I64">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+				<port id="1" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I32">
+				<port id="2" precision="FP32" names="365,emb">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="124" name="Constant_38197" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="125" name="__module.model.gpt_neox.layers.0.attention/aten::index/Gather" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="103" name="__module.model.gpt_neox.rotary_emb/aten::cos/Cos" type="Cos" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="356">
+				<port id="1" precision="FP32" names="366,368,370,373,cos.1,cos.3,cos.5">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="126" name="205" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="201162306" size="8" />
-			<output>
-				<port id="0" precision="I64" names="205" />
-			</output>
-		</layer>
-		<layer id="127" name="__module.model.gpt_neox.layers.0.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1">
+		<layer id="104" name="__module.model.gpt_neox.layers.0.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -1824,7 +1465,7 @@
 				<port id="1" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="357,cos.3">
+				<port id="2" precision="FP32" names="1082,1194,1306,1418,1530,1642,1754,1866,1978,2090,2202,2314,2426,2538,2650,2762,2874,2986,3098,3210,3322,3434,3546,3658,3770,3882,407,522,634,746,858,970,cos,cos.11,cos.13,cos.15,cos.17,cos.19,cos.21,cos.23,cos.25,cos.27,cos.29,cos.31,cos.33,cos.35,cos.37,cos.39,cos.41,cos.43,cos.45,cos.47,cos.49,cos.51,cos.53,cos.55,cos.57,cos.59,cos.61,cos.63,cos.65,cos.67,cos.7,cos.9">
 					<dim>-1</dim>
 					<dim>1</dim>
 					<dim>-1</dim>
@@ -1832,7 +1473,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="128" name="__module.model.gpt_neox.layers.0.attention/aten::mul/Multiply" type="Multiply" version="opset1">
+		<layer id="105" name="__module.model.gpt_neox.layers.0.attention/aten::mul/Multiply" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -1849,7 +1490,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="361">
+				<port id="2" precision="FP32" names="409">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -1857,178 +1498,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="129" name="Constant_269051" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="130" name="Constant_269050" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="106" name="Constant_922" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="131" name="__module.model.gpt_neox.layers.0.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="132" name="207" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64" names="207" />
-			</output>
-		</layer>
-		<layer id="133" name="__module.model.gpt_neox.layers.0.attention/aten::size/Constant_5" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="107" name="Constant_924" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="134" name="__module.model.gpt_neox.layers.0.attention/aten::size/Gather_5" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="362,367" />
-			</output>
-		</layer>
-		<layer id="135" name="Constant_58" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="201162298" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64" names="220" />
-			</output>
-		</layer>
-		<layer id="136" name="__module.model.gpt_neox.layers.0.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" />
-			</output>
-		</layer>
-		<layer id="137" name="__module.model.gpt_neox.layers.0.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-			</input>
-			<output>
-				<port id="1" precision="I64" names="364,365,369,370" />
-			</output>
-		</layer>
-		<layer id="138" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Constant_1" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="139" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="140" name="Constant_269049" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="108" name="Constant_926" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="141" name="ScatterUpdate_269052" type="ScatterUpdate" version="opset3">
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="142" name="Constant_269055" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="143" name="Constant_269058" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
 			</output>
 		</layer>
-		<layer id="144" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="109" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Slice_2" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -2037,17 +1531,20 @@
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="371,x2.1">
+				<port id="5" precision="FP32" names="419,x2.1">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -2055,32 +1552,10 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="145" name="Constant_299073_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="146" name="Constant_299073" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
+		<layer id="110" name="Constant_37292" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>1</dim>
@@ -2088,7 +1563,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="147" name="__module.model.gpt_neox.layers.0.attention/aten::neg/Multiply" type="Multiply" version="opset1">
+		<layer id="111" name="__module.model.gpt_neox.layers.0.attention/aten::neg/Multiply" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -2105,7 +1580,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="372">
+				<port id="2" precision="FP32" names="420">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -2113,30 +1588,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="148" name="Constant_269103" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="112" name="Constant_819" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="149" name="Constant_269108" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="113" name="Constant_863" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="114" name="Constant_865" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="150" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="115" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Slice_3" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -2145,17 +1621,20 @@
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="366">
+				<port id="5" precision="FP32" names="414">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -2163,7 +1642,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="151" name="__module.model.gpt_neox.layers.0.attention/aten::cat/Concat" type="Concat" version="opset1">
+		<layer id="116" name="__module.model.gpt_neox.layers.0.attention/aten::cat/Concat" type="Concat" version="opset1">
 			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
@@ -2180,7 +1659,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="374">
+				<port id="2" precision="FP32" names="422">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -2188,107 +1667,23 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="152" name="self.model.gpt_neox.layers.0.attention.rotary_emb.sin_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201162356" size="81920" />
-			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.0.attention.rotary_emb.sin_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="153" name="self.model.gpt_neox.layers.0.attention.rotary_emb.sin_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="154" name="__module.model.gpt_neox.layers.0.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="155" name="__module.model.gpt_neox.layers.0.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="156" name="__module.model.gpt_neox.layers.0.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="117" name="__module.model.gpt_neox.rotary_emb/aten::sin/Sin" type="Sin" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="351,354,sin.1">
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="157" name="Constant_38200" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="158" name="__module.model.gpt_neox.layers.0.attention/aten::index/Gather_1" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="359">
+				<port id="1" precision="FP32" names="367,369,371,374,sin.1,sin.3,sin.5">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="159" name="__module.model.gpt_neox.layers.0.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1">
+		<layer id="118" name="__module.model.gpt_neox.layers.0.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -2298,7 +1693,7 @@
 				<port id="1" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="360,sin.3">
+				<port id="2" precision="FP32" names="1083,1195,1307,1419,1531,1643,1755,1867,1979,2091,2203,2315,2427,2539,2651,2763,2875,2987,3099,3211,3323,3435,3547,3659,3771,3883,408,523,635,747,859,971,sin,sin.11,sin.13,sin.15,sin.17,sin.19,sin.21,sin.23,sin.25,sin.27,sin.29,sin.31,sin.33,sin.35,sin.37,sin.39,sin.41,sin.43,sin.45,sin.47,sin.49,sin.51,sin.53,sin.55,sin.57,sin.59,sin.61,sin.63,sin.65,sin.67,sin.7,sin.9">
 					<dim>-1</dim>
 					<dim>1</dim>
 					<dim>-1</dim>
@@ -2306,7 +1701,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="160" name="__module.model.gpt_neox.layers.0.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
+		<layer id="119" name="__module.model.gpt_neox.layers.0.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -2323,7 +1718,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="375">
+				<port id="2" precision="FP32" names="423">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -2331,7 +1726,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="161" name="__module.model.gpt_neox.layers.0.attention/aten::add/Add" type="Add" version="opset1">
+		<layer id="120" name="__module.model.gpt_neox.layers.0.attention/aten::add/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -2348,7 +1743,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="376">
+				<port id="2" precision="FP32" names="424">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -2356,41 +1751,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="162" name="Constant_269154" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="121" name="Constant_774" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="163" name="Constant_269157" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="122" name="Constant_776" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="164" name="Constant_269160" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="123" name="Constant_778" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="165" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="124" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Slice_4" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -2399,17 +1784,20 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="338">
+				<port id="5" precision="FP32" names="404">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -2417,7 +1805,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="166" name="__module.model.gpt_neox.layers.0.attention/aten::cat/Concat_1" type="Concat" version="opset1">
+		<layer id="125" name="__module.model.gpt_neox.layers.0.attention/aten::cat/Concat_1" type="Concat" version="opset1">
 			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
@@ -2434,7 +1822,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="394,404,query.3,query.5">
+				<port id="2" precision="FP32" names="442">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -2442,205 +1830,237 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="167" name="Constant_269166" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="126" name="Constant_52551" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
+				<port id="0" precision="FP32" />
 			</output>
 		</layer>
-		<layer id="168" name="Constant_269169" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="127" name="Constant_52544" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="169" name="Constant_269172" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="128" name="Constant_52546" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="170" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="129" name="Constant_52548" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="130" name="Concat_52549" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="4" precision="I64">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="131" name="Broadcast_52552" type="Broadcast" version="opset3">
+			<data mode="numpy" />
+			<input>
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="339,k.1">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="171" name="__module.model.gpt_neox.layers.0.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="132" name="ReadValue_50183" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.0.keypresent.0.key" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
 				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP32" names="past_key_values.0.key">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
+			</output>
+		</layer>
+		<layer id="133" name="Constant_47778" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64" />
+			</output>
+		</layer>
+		<layer id="134" name="Gather_47779" type="Gather" version="opset8">
+			<data batch_dims="0" />
+			<input>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I32">
 					<dim>-1</dim>
-					<dim>20</dim>
 				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="377">
+				<port id="3" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="172" name="Constant_269176" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="135" name="Constant_728" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="173" name="Constant_269175" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="136" name="Constant_730" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="174" name="__module.model.gpt_neox.layers.0.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="137" name="Constant_732" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="138" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Slice_5" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
-			</input>
-			<output>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="175" name="__module.model.gpt_neox.layers.0.attention/aten::size/Constant_7" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="176" name="__module.model.gpt_neox.layers.0.attention/aten::size/Gather_7" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="378,383" />
+				<port id="5" precision="FP32" names="397">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="177" name="__module.model.gpt_neox.layers.0.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
+		<layer id="139" name="Constant_741" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="2" precision="I64" />
+				<port id="0" precision="I64" names="398">
+					<dim>4</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="178" name="__module.model.gpt_neox.layers.0.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1">
+		<layer id="140" name="__module.model.gpt_neox.layers.0.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
 			<input>
-				<port id="0" precision="I64" />
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>4</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64" names="380,381,385,386" />
+				<port id="2" precision="FP32" names="399,key.1">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="179" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="141" name="Constant_784" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="180" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="142" name="Constant_786" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="181" name="Constant_269174" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="143" name="Constant_788" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="182" name="ScatterUpdate_269177" type="ScatterUpdate" version="opset3">
+		<layer id="144" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Slice_6" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -2648,40 +2068,72 @@
 				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="5" precision="FP32" names="405,k.1">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="183" name="Constant_269180" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="145" name="__module.model.gpt_neox.layers.0.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="425">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="146" name="Constant_1065" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="184" name="Constant_269183" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="147" name="Constant_1067" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="148" name="Constant_1069" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="185" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="149" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Slice_7" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -2690,17 +2142,20 @@
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="387,x2.3">
+				<port id="5" precision="FP32" names="435,x2.3">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -2708,32 +2163,10 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="186" name="Constant_299074_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
+		<layer id="150" name="Constant_37293" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="187" name="Constant_299074" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>1</dim>
@@ -2741,7 +2174,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="188" name="__module.model.gpt_neox.layers.0.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
+		<layer id="151" name="__module.model.gpt_neox.layers.0.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -2758,7 +2191,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="388">
+				<port id="2" precision="FP32" names="436">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -2766,30 +2199,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="189" name="Constant_269228" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="152" name="Constant_962" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="190" name="Constant_269233" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="153" name="Constant_1006" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="154" name="Constant_1008" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="191" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="155" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Slice_8" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -2798,17 +2232,20 @@
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="382">
+				<port id="5" precision="FP32" names="430">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -2816,7 +2253,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="192" name="__module.model.gpt_neox.layers.0.attention/aten::cat/Concat_2" type="Concat" version="opset1">
+		<layer id="156" name="__module.model.gpt_neox.layers.0.attention/aten::cat/Concat_2" type="Concat" version="opset1">
 			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
@@ -2833,7 +2270,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="390">
+				<port id="2" precision="FP32" names="438">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -2841,7 +2278,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="193" name="__module.model.gpt_neox.layers.0.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
+		<layer id="157" name="__module.model.gpt_neox.layers.0.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -2858,7 +2295,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="391">
+				<port id="2" precision="FP32" names="439">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -2866,7 +2303,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="194" name="__module.model.gpt_neox.layers.0.attention/aten::add/Add_1" type="Add" version="opset1">
+		<layer id="158" name="__module.model.gpt_neox.layers.0.attention/aten::add/Add_1" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -2883,7 +2320,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="392">
+				<port id="2" precision="FP32" names="440">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -2891,41 +2328,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="195" name="Constant_269278" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="159" name="Constant_794" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="196" name="Constant_269281" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="160" name="Constant_796" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="197" name="Constant_269284" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="161" name="Constant_798" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="198" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="162" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Slice_9" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -2934,17 +2361,20 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="340">
+				<port id="5" precision="FP32" names="406">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -2952,7 +2382,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="199" name="__module.model.gpt_neox.layers.0.attention/aten::cat/Concat_3" type="Concat" version="opset1">
+		<layer id="163" name="__module.model.gpt_neox.layers.0.attention/aten::cat/Concat_3" type="Concat" version="opset1">
 			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
@@ -2969,7 +2399,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="396">
+				<port id="2" precision="FP32" names="444,key_states">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -2977,7 +2407,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="200" name="__module.model.gpt_neox.layers.0.attention/aten::to/Convert_1" type="Concat" version="opset1">
+		<layer id="164" name="__module.model.gpt_neox.layers.0.attention/aten::cat/Concat_4" type="Concat" version="opset1">
 			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
@@ -3002,49 +2432,37 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="201" name="Constant_337442_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="202" name="Constant_337442" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
+		<layer id="165" name="Constant_52563" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="1" precision="FP32" />
+				<port id="0" precision="FP32" />
 			</output>
 		</layer>
-		<layer id="203" name="Constant_337435" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
+		<layer id="166" name="Constant_52556" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="204" name="Constant_337437" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
+		<layer id="167" name="Constant_52558" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="205" name="Constant_337439" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
+		<layer id="168" name="Constant_52560" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="206" name="Concat_337440" type="Concat" version="opset1">
+		<layer id="169" name="Concat_52561" type="Concat" version="opset1">
 			<data axis="0" />
 			<input>
 				<port id="0" precision="I64">
@@ -3066,7 +2484,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="207" name="Broadcast_337443" type="Broadcast" version="opset3">
+		<layer id="170" name="Broadcast_52564" type="Broadcast" version="opset3">
 			<data mode="numpy" />
 			<input>
 				<port id="0" precision="FP32" />
@@ -3083,7 +2501,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="208" name="ReadValue_330427" type="ReadValue" version="opset6">
+		<layer id="171" name="ReadValue_50185" type="ReadValue" version="opset6">
 			<data variable_id="past_key_values.0.valuepresent.0.value" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
 				<port id="0" precision="FP32">
@@ -3102,13 +2520,13 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="209" name="Constant_323388" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
+		<layer id="172" name="Constant_47781" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="210" name="Gather_323389" type="Gather" version="opset8">
+		<layer id="173" name="Gather_47782" type="Gather" version="opset8">
 			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="FP32">
@@ -3131,41 +2549,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="211" name="Constant_269290" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="174" name="Constant_746" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="212" name="Constant_269293" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="175" name="Constant_748" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="213" name="Constant_269296" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="176" name="Constant_750" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="214" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="177" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Slice_10" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -3174,17 +2582,20 @@
 					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="334">
+				<port id="5" precision="FP32" names="400">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -3192,15 +2603,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="215" name="Constant_251" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="178" name="Constant_759" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64" names="335">
+				<port id="0" precision="I64" names="401">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="216" name="__module.model.gpt_neox.layers.0.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
+		<layer id="179" name="__module.model.gpt_neox.layers.0.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -3213,7 +2624,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="336">
+				<port id="2" precision="FP32" names="402">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -3221,7 +2632,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="217" name="__module.model.gpt_neox.layers.0.attention/aten::cat/Concat_5" type="Concat" version="opset1">
+		<layer id="180" name="__module.model.gpt_neox.layers.0.attention/aten::cat/Concat_5" type="Concat" version="opset1">
 			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
@@ -3246,533 +2657,443 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="218" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Slice_11" type="Const" version="opset1">
-			<data element_type="boolean" shape="1, 1, 2048, 2048" offset="201244276" size="4194304" />
+		<layer id="181" name="__module.model.gpt_neox/aten::triu/Constant" type="Const" version="opset1">
+			<data element_type="i32" shape="" offset="0" size="4" />
 			<output>
-				<port id="0" precision="BOOL" names="413">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="219" name="Constant_269300" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="220" name="Constant_269299" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162298" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
+				<port id="0" precision="I32" />
 			</output>
 		</layer>
-		<layer id="221" name="__module.model.gpt_neox.layers.0.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3">
+		<layer id="182" name="ShapeOf_38688" type="ShapeOf" version="opset3">
 			<data output_type="i64" />
 			<input>
-				<port id="0" precision="FP32">
+				<port id="0" precision="I64">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>2</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="222" name="221" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="205438604" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="183" name="Constant_38689" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="I64" names="221" />
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="223" name="__module.model.gpt_neox.layers.0.attention/aten::size/Constant_8" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="184" name="Constant_38690" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="224" name="__module.model.gpt_neox.layers.0.attention/aten::size/Gather_8" type="Gather" version="opset8">
+		<layer id="185" name="Gather_38691" type="Gather" version="opset8">
 			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>2</dim>
 				</port>
 				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="408" />
+				<port id="3" precision="I64" names="309,326" />
 			</output>
 		</layer>
-		<layer id="225" name="__module.model.gpt_neox.layers.0.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="186" name="Convert_38694" type="Convert" version="opset1">
+			<data destination_type="i32" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
+				<port id="0" precision="I64" />
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
+				<port id="1" precision="I32" />
 			</output>
 		</layer>
-		<layer id="226" name="__module.model.gpt_neox.layers.0.attention/aten::size/Constant_9" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="187" name="__module.model.gpt_neox/aten::triu/Constant_2" type="Const" version="opset1">
+			<data element_type="i32" shape="" offset="201116160" size="4" />
 			<output>
 				<port id="0" precision="I32" />
 			</output>
 		</layer>
-		<layer id="227" name="__module.model.gpt_neox.layers.0.attention/aten::size/Gather_9" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="188" name="__module.model.gpt_neox/aten::triu/Range" type="Range" version="opset4">
+			<data output_type="i32" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
+				<port id="0" precision="I32" />
+				<port id="1" precision="I32" />
 				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="406" />
+				<port id="3" precision="I32">
+					<dim>-1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="228" name="__module.model.gpt_neox.layers.0.attention/aten::sub/Subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="189" name="__module.model.gpt_neox/aten::triu/Unsqueeze" type="Unsqueeze" version="opset1">
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
+				<port id="0" precision="I32">
+					<dim>-1</dim>
+				</port>
+				<port id="1" precision="I32" />
 			</input>
 			<output>
-				<port id="2" precision="I64" names="410,411" />
+				<port id="2" precision="I32">
+					<dim>1</dim>
+					<dim>-1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="229" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Constant_4" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="190" name="__module.model.gpt_neox/aten::triu/Convert" type="Const" version="opset1">
+			<data element_type="i32" shape="" offset="201116160" size="4" />
 			<output>
 				<port id="0" precision="I32" />
 			</output>
 		</layer>
-		<layer id="230" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1">
+		<layer id="191" name="ShapeOf_38682" type="ShapeOf" version="opset3">
+			<data output_type="i64" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
+				<port id="0" precision="I32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="1" precision="I64">
+					<dim>2</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="231" name="Constant_269298" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="192" name="Constant_38683" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
+				<port id="0" precision="I64" />
+			</output>
+		</layer>
+		<layer id="193" name="Constant_38684" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="232" name="ScatterUpdate_269301" type="ScatterUpdate" version="opset3">
+		<layer id="194" name="Gather_38685" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
+					<dim>2</dim>
 				</port>
+				<port id="1" precision="I64" />
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
-				</port>
+				<port id="3" precision="I64" names="301,308" />
 			</output>
 		</layer>
-		<layer id="233" name="Constant_269302" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="195" name="Convert_38699" type="Convert" version="opset1">
+			<data destination_type="i32" />
+			<input>
+				<port id="0" precision="I64" />
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
+				<port id="1" precision="I32" />
 			</output>
 		</layer>
-		<layer id="234" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1">
+		<layer id="196" name="__module.model.gpt_neox/aten::triu/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64" />
+				<port id="0" precision="I32" />
 				<port id="1" precision="I32" />
 			</input>
 			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
+				<port id="2" precision="I32" />
 			</output>
 		</layer>
-		<layer id="235" name="ScatterUpdate_269303" type="ScatterUpdate" version="opset3">
+		<layer id="197" name="__module.model.gpt_neox/aten::triu/Range_1" type="Range" version="opset4">
+			<data output_type="i32" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
+				<port id="0" precision="I32" />
+				<port id="1" precision="I32" />
+				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
+				<port id="3" precision="I32">
+					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="236" name="Constant_269306" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438612" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="198" name="__module.model.gpt_neox/aten::triu/Unsqueeze_1" type="Unsqueeze" version="opset1">
+			<input>
+				<port id="0" precision="I32">
+					<dim>-1</dim>
+				</port>
+				<port id="1" precision="I32" />
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="2" precision="I32">
+					<dim>-1</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="237" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="199" name="__module.model.gpt_neox/aten::triu/GreaterEqual" type="GreaterEqual" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
+				<port id="0" precision="I32">
 					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>3</dim>
+					<dim>-1</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>3</dim>
+				<port id="1" precision="I32">
+					<dim>-1</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="BOOL" names="414">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="2" precision="BOOL">
+					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>2048</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="238" name="Constant_269379" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="200" name="__module.model.gpt_neox/aten::full/Convert" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="201116164" size="4" />
+			<output>
+				<port id="0" precision="FP32" />
+			</output>
+		</layer>
+		<layer id="201" name="Constant_38686" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="239" name="Constant_269380" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="202" name="Reshape_38687" type="Reshape" version="opset1">
+			<data special_zero="false" />
+			<input>
+				<port id="0" precision="I64" />
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="240" name="Constant_269376" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="203" name="Constant_38692" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="241" name="Constant_269375" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="204" name="Reshape_38693" type="Reshape" version="opset1">
+			<data special_zero="false" />
+			<input>
+				<port id="0" precision="I64" />
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I32">
+				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="242" name="ScatterUpdate_269381" type="ScatterUpdate" version="opset3">
+		<layer id="205" name="__module.model.gpt_neox/prim::ListConstruct" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
+			</input>
+			<output>
 				<port id="2" precision="I64">
-					<dim>1</dim>
+					<dim>2</dim>
 				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
+			</output>
+		</layer>
+		<layer id="206" name="__module.model.gpt_neox/aten::full/Broadcast" type="Broadcast" version="opset3">
+			<data mode="numpy" />
+			<input>
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
+					<dim>2</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="312">
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="243" name="Constant_269384" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="207" name="__module.model.gpt_neox/aten::triu/ConvertLike" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
+				<port id="0" precision="FP32" />
 			</output>
 		</layer>
-		<layer id="244" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="208" name="__module.model.gpt_neox/aten::triu/Select" type="Select" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>-1</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>-1</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
+				<port id="2" precision="FP32" />
 			</input>
 			<output>
-				<port id="4" precision="BOOL" names="415,416">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="3" precision="FP32" names="313_1">
 					<dim>-1</dim>
 					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="245" name="__module.model.gpt_neox.layers.0.attention/aten::where/Convert_1_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
+		<layer id="209" name="__module.model.gpt_neox/aten::arange/Constant" type="Const" version="opset1">
+			<data element_type="i32" shape="" offset="0" size="4" />
 			<output>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="I32" />
 			</output>
 		</layer>
-		<layer id="246" name="__module.model.gpt_neox.layers.0.attention/aten::where/Convert_1" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
+		<layer id="210" name="__module.model.gpt_neox/aten::arange/Constant_1" type="Const" version="opset1">
+			<data element_type="i32" shape="" offset="201116160" size="4" />
 			<output>
-				<port id="1" precision="FP32" />
+				<port id="0" precision="I32" />
 			</output>
 		</layer>
-		<layer id="247" name="__module.model.gpt_neox.layers.0.attention/aten::full/Convert" type="Const" version="opset1">
-			<data element_type="f32" shape="" offset="205438636" size="4" />
+		<layer id="211" name="__module.model.gpt_neox/aten::arange/Range" type="Range" version="opset4">
+			<data output_type="f32" />
+			<input>
+				<port id="0" precision="I32" />
+				<port id="1" precision="I64" />
+				<port id="2" precision="I32" />
+			</input>
 			<output>
-				<port id="0" precision="FP32" names="403" />
+				<port id="3" precision="FP32">
+					<dim>-1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="248" name="__module.model.gpt_neox.layers.0.attention/aten::where/Select" type="Select" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="212" name="__module.model.gpt_neox/aten::arange/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="i64" />
 			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
 				</port>
-				<port id="1" precision="FP32" />
-				<port id="2" precision="FP32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="417,causal_mask.1">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="1" precision="I64" names="314">
 					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="249" name="Constant_299076_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="205438640" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="250" name="Constant_299076" type="Convert" version="opset1">
+		<layer id="213" name="__module.model.gpt_neox/aten::gt/Convert" type="Convert" version="opset1">
 			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="I64">
+					<dim>-1</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="251" name="__module.model.gpt_neox/aten::size/ShapeOf" type="ShapeOf" version="opset3">
+		<layer id="214" name="ShapeOf_38704" type="ShapeOf" version="opset3">
 			<data output_type="i64" />
 			<input>
-				<port id="0" precision="I64">
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="I64">
-					<dim>2</dim>
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="252" name="Constant_254329" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="215" name="Constant_38705" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="201116168" size="8" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="253" name="__module.model.gpt_neox/aten::size/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="216" name="Constant_38706" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="254" name="__module.model.gpt_neox/aten::size/Gather" type="Gather" version="opset8">
+		<layer id="217" name="Gather_38707" type="Gather" version="opset8">
 			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="I64">
-					<dim>2</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
+					<dim>4</dim>
 				</port>
-				<port id="2" precision="I32" />
+				<port id="1" precision="I64" />
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="296">
-					<dim>1</dim>
-				</port>
+				<port id="3" precision="I64" names="303" />
 			</output>
 		</layer>
-		<layer id="255" name="__module.model.gpt_neox.layers.0.attention/prim::ListConstruct/Reshape_0_1" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="205438642" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="218" name="__module.model.gpt_neox/aten::add/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="I64" />
+				<port id="1" precision="I64" />
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
+				<port id="2" precision="I64" names="305" />
 			</output>
 		</layer>
-		<layer id="256" name="__module.model.gpt_neox/prim::ListConstruct/Concat" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="219" name="__module.model.gpt_neox/aten::arange/Constant_2" type="Const" version="opset1">
+			<data element_type="i32" shape="" offset="201116160" size="4" />
+			<output>
+				<port id="0" precision="I32" />
+			</output>
+		</layer>
+		<layer id="220" name="__module.model.gpt_neox/aten::arange/Range_1" type="Range" version="opset4">
+			<data output_type="f32" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
+				<port id="0" precision="I64" />
+				<port id="1" precision="I64" />
+				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="2" precision="I64" names="297">
+				<port id="3" precision="FP32" names="307,cache_position">
+					<dim>-1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="221" name="Constant_280" type="Const" version="opset1">
+			<data element_type="i64" shape="2" offset="201116176" size="16" />
+			<output>
+				<port id="0" precision="I64" names="315">
 					<dim>2</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="257" name="__module.model.gpt_neox/aten::view/Reshape" type="Reshape" version="opset1">
+		<layer id="222" name="__module.model.gpt_neox/aten::reshape/Reshape" type="Reshape" version="opset1">
 			<data special_zero="false" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>-1</dim>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
 				</port>
 				<port id="1" precision="I64">
@@ -3780,186 +3101,183 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64" names="298,299,attention_mask.3">
-					<dim>-1</dim>
+				<port id="2" precision="FP32" names="316">
 					<dim>-1</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="258" name="__module.model.gpt_neox/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1">
+		<layer id="223" name="__module.model.gpt_neox/aten::gt/Greater" type="Greater" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
+				</port>
+				<port id="1" precision="FP32">
 					<dim>-1</dim>
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="I64" names="300">
+				<port id="2" precision="BOOL" names="317">
 					<dim>-1</dim>
-					<dim>1</dim>
 					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="259" name="__module.model.gpt_neox/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1">
+		<layer id="224" name="__module.model.gpt_neox/aten::mul_/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<input>
-				<port id="0" precision="I64">
+				<port id="0" precision="BOOL">
 					<dim>-1</dim>
-					<dim>1</dim>
 					<dim>-1</dim>
 				</port>
-				<port id="1" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="I64" names="301,302,attention_mask.5">
+				<port id="1" precision="FP32">
 					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="260" name="__module.model.gpt_neox/aten::to/Convert" type="Convert" version="opset1">
-			<data destination_type="f32" />
+		<layer id="225" name="__module.model.gpt_neox/aten::mul_/Multiply" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+				</port>
+				<port id="1" precision="FP32">
 					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>-1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="303">
+				<port id="2" precision="FP32" names="313,causal_mask.1">
 					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="261" name="Constant_299075_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="205438640" size="2" />
+		<layer id="226" name="208" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
+				<port id="0" precision="I64" names="208" />
 			</output>
 		</layer>
-		<layer id="262" name="Constant_299075" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="227" name="__module.model.gpt_neox/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1">
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
+				<port id="1" precision="I64" />
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="319">
 					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="263" name="__module.model.gpt_neox/aten::rsub/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="228" name="__module.model.gpt_neox/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
 					<dim>1</dim>
 					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
+				<port id="1" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
+				<port id="2" precision="FP32" names="320,321,322">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="264" name="__module.model.gpt_neox/aten::rsub/Subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+		<layer id="229" name="Constant_38709" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
+			</output>
+		</layer>
+		<layer id="230" name="Constant_38710" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64" />
+			</output>
+		</layer>
+		<layer id="231" name="Gather_38711" type="Gather" version="opset8">
+			<data batch_dims="0" />
+			<input>
+				<port id="0" precision="I64">
+					<dim>2</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>1</dim>
-					<dim>-1</dim>
 				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="304">
-					<dim>-1</dim>
-					<dim>1</dim>
+				<port id="3" precision="I64" names="310">
 					<dim>1</dim>
-					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="265" name="Constant_299077" type="Const" version="opset1">
-			<data element_type="f32" shape="1, 1, 1, 1" offset="205438636" size="4" />
+		<layer id="232" name="Constant_32663" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="FP32">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="233" name="Constant_32665" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="234" name="Constant_32667" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="266" name="__module.model.gpt_neox/aten::mul/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="235" name="__module.model.gpt_neox/prim::ListConstruct_1" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>-1</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="305,attention_mask">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
+				<port id="4" precision="I64">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="267" name="Add_290338" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="236" name="__module.model.gpt_neox/aten::expand/Broadcast" type="Broadcast" version="opset3">
+			<data mode="bidirectional" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>1</dim>
@@ -3967,15 +3285,12 @@
 					<dim>-1</dim>
 					<dim>-1</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="324,327,328,329,337,338,339,342_1,343_1,344_1,causal_mask.5">
 					<dim>-1</dim>
 					<dim>1</dim>
 					<dim>-1</dim>
@@ -3983,12 +3298,45 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="268" name="ShapeOf_290339" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
+		<layer id="237" name="Constant_26934" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116192" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="238" name="SliceAssign_494/Reshape_3" type="Reshape" version="opset1">
+			<data special_zero="false" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="239" name="Constant_21431" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64" />
+			</output>
+		</layer>
+		<layer id="240" name="SliceAssign_494/ShapeOf" type="ShapeOf" version="opset3">
+			<data output_type="i64" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
 					<dim>-1</dim>
 				</port>
@@ -3999,49 +3347,154 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="269" name="Constant_254326" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="241" name="SliceAssign_494/ReduceProd" type="ReduceProd" version="opset1">
+			<data keep_dims="false" />
+			<input>
+				<port id="0" precision="I64">
+					<dim>4</dim>
+				</port>
+				<port id="1" precision="I64" />
+			</input>
+			<output>
+				<port id="2" precision="I64" />
+			</output>
+		</layer>
+		<layer id="242" name="Constant_21432" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64" />
+			</output>
+		</layer>
+		<layer id="243" name="SliceAssign_494/Range" type="Range" version="opset4">
+			<data output_type="i64" />
+			<input>
+				<port id="0" precision="I64" />
+				<port id="1" precision="I64" />
+				<port id="2" precision="I64" />
+			</input>
+			<output>
+				<port id="3" precision="I64">
+					<dim>-1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="244" name="SliceAssign_494/Reshape" type="Reshape" version="opset1">
+			<data special_zero="false" />
+			<input>
+				<port id="0" precision="I64">
+					<dim>-1</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>4</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="I64">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="245" name="Constant_483" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="270" name="__module.model.gpt_neox.layers.0.attention/aten::size/Constant_10" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="246" name="Constant_486" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="271" name="__module.model.gpt_neox.layers.0.attention/aten::size/Gather_10" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="247" name="SliceAssign_494/Slice_2" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I32" />
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="3" precision="I64" names="401">
+				<port id="5" precision="I64">
+					<dim>-1</dim>
 					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="272" name="__module.model.gpt_neox.layers.0.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="248" name="Constant_21434" type="Const" version="opset1">
+			<data element_type="i64" shape="2" offset="201116176" size="16" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>2</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="249" name="SliceAssign_494/Reshape_4" type="Reshape" version="opset1">
+			<data special_zero="false" />
 			<input>
+				<port id="0" precision="I64">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>2</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="I64">
+					<dim>-1</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="250" name="Constant_366" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="251" name="Constant_369" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
+			</output>
+		</layer>
+		<layer id="252" name="__module.model.gpt_neox/aten::slice/Slice_5" type="Slice" version="opset8">
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
+				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
@@ -4051,48 +3504,253 @@
 				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64" names="418">
-					<dim>4</dim>
+				<port id="5" precision="FP32" names="330,340,345_1">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="273" name="__module.model.gpt_neox.layers.0.attention/aten::expand/Abs" type="Abs" version="opset1">
+		<layer id="253" name="__module.model.gpt_neox/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1">
 			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
+				<port id="1" precision="I64" />
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="I64" names="332">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="274" name="Maximum_290340" type="Maximum" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="254" name="202" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="201116168" size="8" />
+			<output>
+				<port id="0" precision="I64" names="202" />
+			</output>
+		</layer>
+		<layer id="255" name="__module.model.gpt_neox/aten::unsqueeze/Unsqueeze_3" type="Unsqueeze" version="opset1">
+			<input>
+				<port id="0" precision="I64">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+				</port>
+				<port id="1" precision="I64" />
+			</input>
+			<output>
+				<port id="2" precision="I64" names="333,334">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="256" name="__module.model.gpt_neox/aten::add/Convert" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<input>
 				<port id="0" precision="I64">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="257" name="__module.model.gpt_neox/aten::add/Add_1" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="335,padding_mask.1">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="258" name="Constant_37294" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="0" size="4" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="259" name="__module.model.gpt_neox/aten::eq/Equal" type="Equal" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="BOOL" names="336,padding_mask">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="260" name="__module.model.gpt_neox/aten::masked_fill/Select" type="Select" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="BOOL">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
+				</port>
+				<port id="1" precision="FP32" />
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="3" precision="FP32" names="341">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="261" name="__module.model.gpt_neox/aten::copy_/ShapeOf" type="ShapeOf" version="opset3">
+			<data output_type="i64" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
+			</output>
+		</layer>
+		<layer id="262" name="__module.model.gpt_neox/aten::copy_/Broadcast" type="Broadcast" version="opset3">
+			<data mode="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
+				</port>
 				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="345">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="275" name="__module.model.gpt_neox.layers.0.attention/aten::add/Add_2" type="Broadcast" version="opset3">
-			<data mode="numpy" />
+		<layer id="263" name="Constant_26936" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116192" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="264" name="SliceAssign_494/Reshape_5" type="Reshape" version="opset1">
+			<data special_zero="false" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="265" name="SliceAssign_494/ScatterNDUpdate" type="ScatterNDUpdate" version="opset4">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>-1</dim>
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="3" precision="FP32">
 					<dim>-1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="266" name="SliceAssign_494/Reshape_6" type="Reshape" version="opset1">
+			<data special_zero="false" />
+			<input>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
 				</port>
 				<port id="1" precision="I64">
@@ -4100,15 +3758,99 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="420">
+				<port id="2" precision="FP32" names="1013,1014,1015,1125,1126,1127,1237,1238,1239,1349,1350,1351,1461,1462,1463,1573,1574,1575,1685,1686,1687,1797,1798,1799,1909,1910,1911,2021,2022,2023,2133,2134,2135,2245,2246,2247,2357,2358,2359,2469,2470,2471,2581,2582,2583,2693,2694,2695,2805,2806,2807,2917,2918,2919,3029,3030,3031,3141,3142,3143,325,3253,3254,3255,3365,3366,3367,3477,3478,3479,3589,3590,3591,3701,3702,3703,3813,3814,3815,3925,3926,3927,453,454,455,565,566,567,677,678,679,789,790,791,901,902,903">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="267" name="Constant_1147" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="268" name="Constant_38714" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="269" name="Reshape_38715" type="Reshape" version="opset1">
+			<data special_zero="false" />
+			<input>
+				<port id="0" precision="I64" />
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="270" name="Add_38718" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="I64" names="452">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="271" name="Constant_1150" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="272" name="__module.model.gpt_neox.layers.0.attention/aten::slice/Slice_14" type="Slice" version="opset8">
+			<input>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="5" precision="FP32" names="456">
 					<dim>-1</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
 					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="276" name="__module.model.gpt_neox.layers.0.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+		<layer id="273" name="__module.model.gpt_neox.layers.0.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
 			<data causal="false" />
 			<input>
 				<port id="0" precision="FP32">
@@ -4131,13 +3873,13 @@
 				</port>
 				<port id="3" precision="FP32">
 					<dim>-1</dim>
-					<dim>-1</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
 					<dim>-1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="421,422,sdpa_result.1,tensor.1">
+				<port id="4" precision="FP32" names="457,attn_output.1">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -4145,15 +3887,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="277" name="Constant_692" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="274" name="__module.model.gpt_neox.layers.0.attention/aten::transpose/Constant" type="Const" version="opset1">
+			<data element_type="i32" shape="4" offset="201116200" size="16" />
 			<output>
-				<port id="0" precision="I64" names="423">
+				<port id="0" precision="I32">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="278" name="__module.model.gpt_neox.layers.0.attention/aten::permute/Transpose_3" type="Transpose" version="opset1">
+		<layer id="275" name="__module.model.gpt_neox.layers.0.attention/aten::transpose/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -4161,12 +3903,12 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="I32">
 					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="424">
+				<port id="2" precision="FP32" names="458">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -4174,18 +3916,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="279" name="__module.model.gpt_neox.layers.0.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438650" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="276" name="Constant_40984" type="Const" version="opset1">
+			<data element_type="i64" shape="3" offset="201116216" size="24" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>3</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="280" name="__module.model.gpt_neox.layers.0.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
+		<layer id="277" name="__module.model.gpt_neox.layers.0.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
 			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -4199,15 +3938,15 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="429">
+				<port id="2" precision="FP32" names="461">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="281" name="self.model.gpt_neox.layers.0.attention.dense.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 2560" offset="205438674" size="6553600" />
+		<layer id="278" name="self.model.gpt_neox.layers.0.attention.dense.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 2560" offset="201116240" size="6553600" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>2560</dim>
@@ -4215,7 +3954,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="282" name="Convert_21434" type="Convert" version="opset1">
+		<layer id="279" name="Convert_75176" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -4230,8 +3969,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="283" name="self.model.gpt_neox.layers.0.attention.dense.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="211992274" size="2560" />
+		<layer id="280" name="self.model.gpt_neox.layers.0.attention.dense.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="207669840" size="2560" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>2560</dim>
@@ -4239,7 +3978,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="284" name="Convert_21437" type="Convert" version="opset1">
+		<layer id="281" name="Convert_75179" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -4254,7 +3993,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="285" name="self.model.gpt_neox.layers.0.attention.dense.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="282" name="self.model.gpt_neox.layers.0.attention.dense.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -4273,8 +4012,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="286" name="self.model.gpt_neox.layers.0.attention.dense.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="211994834" size="5120" />
+		<layer id="283" name="self.model.gpt_neox.layers.0.attention.dense.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="207672400" size="5120" />
 			<output>
 				<port id="0" precision="FP16">
 					<dim>2560</dim>
@@ -4282,7 +4021,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="287" name="self.model.gpt_neox.layers.0.attention.dense.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="284" name="self.model.gpt_neox.layers.0.attention.dense.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -4301,11 +4040,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="288" name="self.model.gpt_neox.layers.0.attention.dense.weight" type="Convert" version="opset1">
+		<layer id="285" name="self.model.gpt_neox.layers.0.attention.dense.weight/fq_weights_1/convert" type="Convert" version="opset1">
 			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
 			<input>
 				<port id="0" precision="FP16">
 					<dim>2560</dim>
@@ -4319,7 +4055,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="289" name="__module.model.gpt_neox.layers.0.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="286" name="__module.model.gpt_neox.layers.0.attention.dense/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -4340,37 +4076,17 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="290" name="Constant_299078_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="211999954" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="291" name="Constant_299078" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="287" name="Constant_37295" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="207677520" size="10240" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="292" name="__module.model.gpt_neox.layers.0.attention.dense/aten::linear/Add" type="Add" version="opset1">
+		<layer id="288" name="__module.model.gpt_neox.layers.0.attention.dense/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -4385,14 +4101,14 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="432,434,input.3">
+				<port id="2" precision="FP32" names="464,466,input.1">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="293" name="__module.model.gpt_neox.layers.0/aten::add/Add" type="Add" version="opset1">
+		<layer id="289" name="__module.model.gpt_neox.layers.0/aten::add/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -4407,14 +4123,14 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="452">
+				<port id="2" precision="FP32" names="484">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="294" name="__module.model.gpt_neox.layers.0/aten::add/Add_1" type="Add" version="opset1">
+		<layer id="290" name="__module.model.gpt_neox.layers.0/aten::add/Add_1" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -4429,22 +4145,22 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="453,455,hidden_states.3">
+				<port id="2" precision="FP32" names="485,487,hidden_states.7">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="295" name="__module.model.gpt_neox.layers.1.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867670" size="4" />
+		<layer id="291" name="__module.model.gpt_neox.layers.1.input_layernorm/aten::layer_norm/Multiply" type="Const" version="opset1">
+			<data element_type="i32" shape="1" offset="128867668" size="4" />
 			<output>
 				<port id="0" precision="I32">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="296" name="__module.model.gpt_neox.layers.1.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
+		<layer id="292" name="__module.model.gpt_neox.layers.1.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
 			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
 			<input>
 				<port id="0" precision="FP32">
@@ -4464,37 +4180,17 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="297" name="Constant_299079_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="212005074" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="298" name="Constant_299079" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="293" name="Constant_37296" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="207687760" size="10240" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="299" name="__module.model.gpt_neox.layers.1.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
+		<layer id="294" name="__module.model.gpt_neox.layers.1.post_attention_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -4516,37 +4212,17 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="300" name="Constant_299080_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="212010194" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="301" name="Constant_299080" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="295" name="Constant_37297" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="207698000" size="10240" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="302" name="__module.model.gpt_neox.layers.1.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+		<layer id="296" name="__module.model.gpt_neox.layers.1.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -4561,15 +4237,15 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="590">
+				<port id="2" precision="FP32" names="585">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="303" name="self.model.gpt_neox.layers.1.mlp.dense_h_to_4h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 2560" offset="212015314" size="26214400" />
+		<layer id="297" name="self.model.gpt_neox.layers.1.mlp.dense_h_to_4h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 2560" offset="207708240" size="26214400" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>10240</dim>
@@ -4577,7 +4253,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="304" name="Convert_21484" type="Convert" version="opset1">
+		<layer id="298" name="Convert_74204" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -4592,8 +4268,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="305" name="self.model.gpt_neox.layers.1.mlp.dense_h_to_4h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 1" offset="238229714" size="10240" />
+		<layer id="299" name="self.model.gpt_neox.layers.1.mlp.dense_h_to_4h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 1" offset="233922640" size="10240" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>10240</dim>
@@ -4601,7 +4277,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="306" name="Convert_21487" type="Convert" version="opset1">
+		<layer id="300" name="Convert_74207" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -4616,7 +4292,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="307" name="self.model.gpt_neox.layers.1.mlp.dense_h_to_4h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="301" name="self.model.gpt_neox.layers.1.mlp.dense_h_to_4h.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -4635,8 +4311,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="308" name="self.model.gpt_neox.layers.1.mlp.dense_h_to_4h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="10240, 1" offset="238239954" size="20480" />
+		<layer id="302" name="self.model.gpt_neox.layers.1.mlp.dense_h_to_4h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="10240, 1" offset="233932880" size="20480" />
 			<output>
 				<port id="0" precision="FP16">
 					<dim>10240</dim>
@@ -4644,7 +4320,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="309" name="self.model.gpt_neox.layers.1.mlp.dense_h_to_4h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="303" name="self.model.gpt_neox.layers.1.mlp.dense_h_to_4h.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -4663,7 +4339,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="310" name="self.model.gpt_neox.layers.1.mlp.dense_h_to_4h.weight" type="Convert" version="opset1">
+		<layer id="304" name="__module.model.gpt_neox.layers.1.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
@@ -4681,7 +4357,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="311" name="__module.model.gpt_neox.layers.1.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="305" name="__module.model.gpt_neox.layers.1.mlp.dense_h_to_4h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -4702,37 +4378,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="312" name="Constant_299081_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 10240" offset="238260434" size="20480" />
+		<layer id="306" name="self.model.gpt_neox.layers.1.mlp.dense_h_to_4h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="10240" offset="233953360" size="20480" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.1.mlp.dense_h_to_4h.bias">
 					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="313" name="Constant_299081" type="Convert" version="opset1">
+		<layer id="307" name="__module.model.gpt_neox.layers.1.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
 			</rt_info>
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16">
 					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="314" name="__module.model.gpt_neox.layers.1.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1">
+		<layer id="308" name="__module.model.gpt_neox.layers.1.mlp.dense_h_to_4h/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -4741,20 +4411,18 @@
 					<dim>10240</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="595">
+				<port id="2" precision="FP32" names="590">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="315" name="__module.model.gpt_neox.layers.1.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
+		<layer id="309" name="__module.model.gpt_neox.layers.1.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
 			<data approximation_mode="ERF" />
 			<input>
 				<port id="0" precision="FP32">
@@ -4764,15 +4432,15 @@
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="596">
+				<port id="1" precision="FP32" names="591">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="316" name="self.model.gpt_neox.layers.1.mlp.dense_4h_to_h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 10240" offset="238280914" size="26214400" />
+		<layer id="310" name="self.model.gpt_neox.layers.1.mlp.dense_4h_to_h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 10240" offset="233973840" size="26214400" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>2560</dim>
@@ -4780,7 +4448,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="317" name="Convert_21494" type="Convert" version="opset1">
+		<layer id="311" name="Convert_74214" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -4795,8 +4463,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="318" name="self.model.gpt_neox.layers.1.mlp.dense_4h_to_h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="264495314" size="2560" />
+		<layer id="312" name="self.model.gpt_neox.layers.1.mlp.dense_4h_to_h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="260188240" size="2560" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>2560</dim>
@@ -4804,7 +4472,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="319" name="Convert_21497" type="Convert" version="opset1">
+		<layer id="313" name="Convert_74217" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -4819,7 +4487,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="320" name="self.model.gpt_neox.layers.1.mlp.dense_4h_to_h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="314" name="self.model.gpt_neox.layers.1.mlp.dense_4h_to_h.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -4838,8 +4506,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="321" name="self.model.gpt_neox.layers.1.mlp.dense_4h_to_h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="264497874" size="5120" />
+		<layer id="315" name="self.model.gpt_neox.layers.1.mlp.dense_4h_to_h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="260190800" size="5120" />
 			<output>
 				<port id="0" precision="FP16">
 					<dim>2560</dim>
@@ -4847,7 +4515,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="322" name="self.model.gpt_neox.layers.1.mlp.dense_4h_to_h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="316" name="self.model.gpt_neox.layers.1.mlp.dense_4h_to_h.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -4866,7 +4534,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="323" name="self.model.gpt_neox.layers.1.mlp.dense_4h_to_h.weight" type="Convert" version="opset1">
+		<layer id="317" name="__module.model.gpt_neox.layers.1.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
@@ -4884,7 +4552,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="324" name="__module.model.gpt_neox.layers.1.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="318" name="__module.model.gpt_neox.layers.1.mlp.dense_4h_to_h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -4905,37 +4573,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="325" name="Constant_299082_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="264502994" size="5120" />
+		<layer id="319" name="self.model.gpt_neox.layers.1.mlp.dense_4h_to_h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="2560" offset="260195920" size="5120" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.1.mlp.dense_4h_to_h.bias">
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="326" name="Constant_299082" type="Convert" version="opset1">
+		<layer id="320" name="__module.model.gpt_neox.layers.1.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
 			</rt_info>
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16">
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="327" name="__module.model.gpt_neox.layers.1.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1">
+		<layer id="321" name="__module.model.gpt_neox.layers.1.mlp.dense_4h_to_h/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -4944,50 +4606,28 @@
 					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="599,input.9">
+				<port id="2" precision="FP32" names="594,input.7">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="328" name="Constant_299083_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="264508114" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="329" name="Constant_299083" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="322" name="Constant_37298" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="260201040" size="10240" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="330" name="__module.model.gpt_neox.layers.1.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
+		<layer id="323" name="__module.model.gpt_neox.layers.1.input_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -5009,37 +4649,17 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="331" name="Constant_299084_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="264513234" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="332" name="Constant_299084" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="324" name="Constant_37299" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="260211280" size="10240" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="333" name="__module.model.gpt_neox.layers.1.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+		<layer id="325" name="__module.model.gpt_neox.layers.1.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -5054,15 +4674,15 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="465">
+				<port id="2" precision="FP32" names="497,hidden_states.5">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="334" name="self.model.gpt_neox.layers.1.attention.query_key_value.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 2560" offset="264518354" size="19660800" />
+		<layer id="326" name="self.model.gpt_neox.layers.1.attention.query_key_value.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 2560" offset="260221520" size="19660800" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>7680</dim>
@@ -5070,7 +4690,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="335" name="Convert_21464" type="Convert" version="opset1">
+		<layer id="327" name="Convert_74835" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -5085,8 +4705,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="336" name="self.model.gpt_neox.layers.1.attention.query_key_value.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 1" offset="284179154" size="7680" />
+		<layer id="328" name="self.model.gpt_neox.layers.1.attention.query_key_value.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 1" offset="279882320" size="7680" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>7680</dim>
@@ -5094,7 +4714,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="337" name="Convert_21467" type="Convert" version="opset1">
+		<layer id="329" name="Convert_74838" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -5109,7 +4729,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="338" name="self.model.gpt_neox.layers.1.attention.query_key_value.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="330" name="self.model.gpt_neox.layers.1.attention.query_key_value.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -5128,8 +4748,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="339" name="self.model.gpt_neox.layers.1.attention.query_key_value.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="7680, 1" offset="284186834" size="15360" />
+		<layer id="331" name="self.model.gpt_neox.layers.1.attention.query_key_value.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="7680, 1" offset="279890000" size="15360" />
 			<output>
 				<port id="0" precision="FP16">
 					<dim>7680</dim>
@@ -5137,7 +4757,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="340" name="self.model.gpt_neox.layers.1.attention.query_key_value.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="332" name="self.model.gpt_neox.layers.1.attention.query_key_value.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -5156,11 +4776,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="341" name="self.model.gpt_neox.layers.1.attention.query_key_value.weight" type="Convert" version="opset1">
+		<layer id="333" name="self.model.gpt_neox.layers.1.attention.query_key_value.weight/fq_weights_1/convert" type="Convert" version="opset1">
 			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
 			<input>
 				<port id="0" precision="FP16">
 					<dim>7680</dim>
@@ -5174,7 +4791,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="342" name="__module.model.gpt_neox.layers.1.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="334" name="__module.model.gpt_neox.layers.1.attention.query_key_value/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -5195,37 +4812,17 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="343" name="Constant_299085_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 7680" offset="284202194" size="15360" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="344" name="Constant_299085" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</input>
+		<layer id="335" name="Constant_37300" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 7680" offset="279905360" size="30720" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="345" name="__module.model.gpt_neox.layers.1.attention.query_key_value/aten::linear/Add" type="Add" version="opset1">
+		<layer id="336" name="__module.model.gpt_neox.layers.1.attention.query_key_value/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -5240,25 +4837,22 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="472,qkv.5">
+				<port id="2" precision="FP32" names="504,qkv.5">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="346" name="__module.model.gpt_neox.layers.1.attention/prim::ListConstruct/Concat" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080154" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="337" name="__module.model.gpt_neox.layers.1.attention/prim::ListConstruct" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201115992" size="32" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="347" name="__module.model.gpt_neox.layers.1.attention/aten::view/Reshape" type="Reshape" version="opset1">
+		<layer id="338" name="__module.model.gpt_neox.layers.1.attention/aten::view/Reshape" type="Reshape" version="opset1">
 			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -5271,7 +4865,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="476,qkv.7">
+				<port id="2" precision="FP32" names="508,qkv.7">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -5279,41 +4873,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="348" name="Constant_269404" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="339" name="Constant_1346" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="349" name="Constant_269407" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="340" name="Constant_1348" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="350" name="Constant_269410" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="341" name="Constant_1350" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="351" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="342" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Slice" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -5322,17 +4906,20 @@
 					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="477">
+				<port id="5" precision="FP32" names="509">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -5340,15 +4927,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="352" name="Constant_804" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="343" name="Constant_1359" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64" names="478">
+				<port id="0" precision="I64" names="510">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="353" name="__module.model.gpt_neox.layers.1.attention/aten::permute/Transpose" type="Transpose" version="opset1">
+		<layer id="344" name="__module.model.gpt_neox.layers.1.attention/aten::permute/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -5361,7 +4948,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="479,query.7">
+				<port id="2" precision="FP32" names="511,query.3">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -5369,41 +4956,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="354" name="Constant_269416" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="345" name="Constant_1400" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="355" name="Constant_269419" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="346" name="Constant_1402" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="356" name="Constant_269422" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="347" name="Constant_1404" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="357" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="348" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Slice_1" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -5412,17 +4989,20 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="486,q.3">
+				<port id="5" precision="FP32" names="518,q.3">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -5430,215 +5010,360 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="358" name="self.model.gpt_neox.layers.1.attention.rotary_emb.cos_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201080346" size="81920" />
+		<layer id="349" name="__module.model.gpt_neox.layers.1.attention/aten::mul/Multiply" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.1.attention.rotary_emb.cos_cached">
-					<dim>2048</dim>
+				<port id="2" precision="FP32" names="524">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="359" name="self.model.gpt_neox.layers.1.attention.rotary_emb.cos_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="350" name="Constant_1558" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="351" name="Constant_1560" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="352" name="Constant_1562" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="353" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Slice_2" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="5" precision="FP32" names="534,x2.5">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="360" name="__module.model.gpt_neox.layers.1.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="354" name="Constant_37301" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="361" name="Constant_269428" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="355" name="__module.model.gpt_neox.layers.1.attention/aten::neg/Multiply" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="535">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="356" name="Constant_1455" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="362" name="Constant_269431" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="357" name="Constant_1499" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="363" name="Constant_269434" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="358" name="Constant_1501" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="364" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="359" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Slice_3" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>240</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="480">
-					<dim>-1</dim>
+				<port id="5" precision="FP32" names="529">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="365" name="Constant_818" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="360" name="__module.model.gpt_neox.layers.1.attention/aten::cat/Concat" type="Concat" version="opset1">
+			<data axis="-1" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64" names="481">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="537">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="366" name="__module.model.gpt_neox.layers.1.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
+		<layer id="361" name="__module.model.gpt_neox.layers.1.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="482,key.7">
+				<port id="2" precision="FP32" names="538">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="367" name="__module.model.gpt_neox.layers.1.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="362" name="__module.model.gpt_neox.layers.1.attention/aten::add/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="539">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="368" name="__module.model.gpt_neox.layers.1.attention/aten::size/Constant_2" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="363" name="Constant_1410" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="369" name="__module.model.gpt_neox.layers.1.attention/aten::size/Gather_2" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
+		<layer id="364" name="Constant_1412" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
+			</output>
+		</layer>
+		<layer id="365" name="Constant_1414" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="3" precision="I64" names="490" />
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="370" name="Constant_337455_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
+		<layer id="366" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Slice_4" type="Slice" version="opset8">
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="FP16" />
+				<port id="5" precision="FP32" names="519">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="371" name="Constant_337455" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="367" name="__module.model.gpt_neox.layers.1.attention/aten::cat/Concat_1" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" />
+				<port id="2" precision="FP32" names="557">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="368" name="Constant_52575" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
+			<output>
+				<port id="0" precision="FP32" />
 			</output>
 		</layer>
-		<layer id="372" name="Constant_337448" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
+		<layer id="369" name="Constant_52568" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="373" name="Constant_337450" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
+		<layer id="370" name="Constant_52570" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="374" name="Constant_337452" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
+		<layer id="371" name="Constant_52572" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="375" name="Concat_337453" type="Concat" version="opset1">
+		<layer id="372" name="Concat_52573" type="Concat" version="opset1">
 			<data axis="0" />
 			<input>
 				<port id="0" precision="I64">
@@ -5660,7 +5385,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="376" name="Broadcast_337456" type="Broadcast" version="opset3">
+		<layer id="373" name="Broadcast_52576" type="Broadcast" version="opset3">
 			<data mode="numpy" />
 			<input>
 				<port id="0" precision="FP32" />
@@ -5677,7 +5402,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="377" name="ReadValue_330429" type="ReadValue" version="opset6">
+		<layer id="374" name="ReadValue_50187" type="ReadValue" version="opset6">
 			<data variable_id="past_key_values.1.keypresent.1.key" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
 				<port id="0" precision="FP32">
@@ -5696,13 +5421,13 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="378" name="Constant_323391" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
+		<layer id="375" name="Constant_47784" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="379" name="Gather_323392" type="Gather" version="opset8">
+		<layer id="376" name="Gather_47785" type="Gather" version="opset8">
 			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="FP32">
@@ -5725,91 +5450,37 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="380" name="__module.model.gpt_neox.layers.1.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="381" name="__module.model.gpt_neox.layers.1.attention/aten::size/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="377" name="Constant_1364" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="382" name="__module.model.gpt_neox.layers.1.attention/aten::size/Gather_3" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="492" />
-			</output>
-		</layer>
-		<layer id="383" name="__module.model.gpt_neox.layers.1.attention/aten::add_/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" names="491,495,496,seq_len.5" />
-			</output>
-		</layer>
-		<layer id="384" name="__module.model.gpt_neox.layers.1.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
 			</output>
 		</layer>
-		<layer id="385" name="__module.model.gpt_neox.layers.1.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="378" name="Constant_1366" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="386" name="__module.model.gpt_neox.layers.1.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="379" name="Constant_1368" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="387" name="__module.model.gpt_neox.layers.1.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="380" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Slice_5" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -5820,60 +5491,103 @@
 				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="499,502,cos.5">
+				<port id="5" precision="FP32" names="512">
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="388" name="Constant_38203" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
+		<layer id="381" name="Constant_1377" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="I64" names="513">
+					<dim>4</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="389" name="__module.model.gpt_neox.layers.1.attention/aten::index/Gather" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="382" name="__module.model.gpt_neox.layers.1.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="505">
+				<port id="2" precision="FP32" names="514,key.5">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="383" name="Constant_1420" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="384" name="Constant_1422" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="385" name="Constant_1424" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="390" name="__module.model.gpt_neox.layers.1.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1">
+		<layer id="386" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Slice_6" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="506,cos.7">
+				<port id="5" precision="FP32" names="520,k.3">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="391" name="__module.model.gpt_neox.layers.1.attention/aten::mul/Multiply" type="Multiply" version="opset1">
+		<layer id="387" name="__module.model.gpt_neox.layers.1.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -5890,7 +5604,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="510">
+				<port id="2" precision="FP32" names="540">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -5898,30 +5612,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="392" name="Constant_269439" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="388" name="Constant_1701" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="393" name="Constant_269438" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="389" name="Constant_1703" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="394" name="__module.model.gpt_neox.layers.1.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="390" name="Constant_1705" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="391" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Slice_7" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -5929,129 +5644,89 @@
 					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="5" precision="FP32" names="550,x2.7">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="395" name="__module.model.gpt_neox.layers.1.attention/aten::size/Constant_5" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="392" name="Constant_37302" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="396" name="__module.model.gpt_neox.layers.1.attention/aten::size/Gather_5" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="511,516" />
-			</output>
-		</layer>
-		<layer id="397" name="__module.model.gpt_neox.layers.1.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" />
-			</output>
-		</layer>
-		<layer id="398" name="__module.model.gpt_neox.layers.1.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-			</input>
-			<output>
-				<port id="1" precision="I64" names="513,514,518,519" />
-			</output>
-		</layer>
-		<layer id="399" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Constant_1" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="400" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
 					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="401" name="Constant_269437" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="402" name="ScatterUpdate_269440" type="ScatterUpdate" version="opset3">
+		<layer id="393" name="__module.model.gpt_neox.layers.1.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="FP32">
+					<dim>1</dim>
 					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
 					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="551">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="403" name="Constant_269443" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="394" name="Constant_1598" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="404" name="Constant_269446" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="395" name="Constant_1642" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="396" name="Constant_1644" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="405" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="397" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Slice_8" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -6060,17 +5735,20 @@
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="520,x2.5">
+				<port id="5" precision="FP32" names="545">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -6078,41 +5756,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="406" name="Constant_299086_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="407" name="Constant_299086" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="408" name="__module.model.gpt_neox.layers.1.attention/aten::neg/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="398" name="__module.model.gpt_neox.layers.1.attention/aten::cat/Concat_2" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -6121,45 +5766,23 @@
 					<dim>10</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="521">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
 					<dim>10</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="409" name="Constant_269491" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="410" name="Constant_269496" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="553">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="411" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="399" name="__module.model.gpt_neox.layers.1.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -6167,43 +5790,40 @@
 					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="515">
+				<port id="2" precision="FP32" names="554">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="412" name="__module.model.gpt_neox.layers.1.attention/aten::cat/Concat" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="400" name="__module.model.gpt_neox.layers.1.attention/aten::add/Add_1" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="523">
+				<port id="2" precision="FP32" names="555">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -6211,61 +5831,37 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="413" name="self.model.gpt_neox.layers.1.attention.rotary_emb.sin_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201162356" size="81920" />
-			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.1.attention.rotary_emb.sin_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="414" name="self.model.gpt_neox.layers.1.attention.rotary_emb.sin_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</input>
+		<layer id="401" name="Constant_1430" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="415" name="__module.model.gpt_neox.layers.1.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="402" name="Constant_1432" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="416" name="__module.model.gpt_neox.layers.1.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="403" name="Constant_1434" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="417" name="__module.model.gpt_neox.layers.1.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="404" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Slice_9" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -6276,61 +5872,21 @@
 				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="500,503,sin.5">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="418" name="Constant_38206" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="419" name="__module.model.gpt_neox.layers.1.attention/aten::index/Gather_1" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="508">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="420" name="__module.model.gpt_neox.layers.1.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="509,sin.7">
+				<port id="5" precision="FP32" names="521">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>60</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="421" name="__module.model.gpt_neox.layers.1.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="405" name="__module.model.gpt_neox.layers.1.attention/aten::cat/Concat_3" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -6340,124 +5896,155 @@
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>60</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="524">
+				<port id="2" precision="FP32" names="559">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="422" name="__module.model.gpt_neox.layers.1.attention/aten::add/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="406" name="__module.model.gpt_neox.layers.1.attention/aten::cat/Concat_4" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="525">
+				<port id="2" precision="FP32" names="present.1.key">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="423" name="Constant_269542" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="407" name="Constant_52587" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
+			<output>
+				<port id="0" precision="FP32" />
+			</output>
+		</layer>
+		<layer id="408" name="Constant_52580" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="424" name="Constant_269545" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="409" name="Constant_52582" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="425" name="Constant_269548" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="410" name="Constant_52584" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="426" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="411" name="Concat_52585" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="4" precision="I64">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="412" name="Broadcast_52588" type="Broadcast" version="opset3">
+			<data mode="numpy" />
+			<input>
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="487">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="427" name="__module.model.gpt_neox.layers.1.attention/aten::cat/Concat_1" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="413" name="ReadValue_50189" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.1.valuepresent.1.value" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP32" names="past_key_values.1.value">
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
+			</output>
+		</layer>
+		<layer id="414" name="Constant_47787" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64" />
+			</output>
+		</layer>
+		<layer id="415" name="Gather_47788" type="Gather" version="opset8">
+			<data batch_dims="0" />
+			<input>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>60</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I32">
+					<dim>-1</dim>
 				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="543,553,query.11,query.9">
+				<port id="3" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -6465,122 +6052,130 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="428" name="Constant_269554" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="416" name="Constant_1382" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="429" name="Constant_269557" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="417" name="Constant_1384" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="430" name="Constant_269560" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="418" name="Constant_1386" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="431" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="419" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Slice_10" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="5" precision="FP32" names="515">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="420" name="Constant_1395" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
+			<output>
+				<port id="0" precision="I64" names="516">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="421" name="__module.model.gpt_neox.layers.1.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="488,k.3">
+				<port id="2" precision="FP32" names="517">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="432" name="__module.model.gpt_neox.layers.1.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="422" name="__module.model.gpt_neox.layers.1.attention/aten::cat/Concat_5" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="526">
+				<port id="2" precision="FP32" names="present.1.value">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="433" name="Constant_269564" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="434" name="Constant_269563" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="423" name="Constant_1777" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="435" name="__module.model.gpt_neox.layers.1.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3">
+		<layer id="424" name="ShapeOf_38783" type="ShapeOf" version="opset3">
 			<data output_type="i64" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
@@ -6589,81 +6184,68 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="436" name="__module.model.gpt_neox.layers.1.attention/aten::size/Constant_7" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="425" name="Constant_38784" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116168" size="8" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="437" name="__module.model.gpt_neox.layers.1.attention/aten::size/Gather_7" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="527,532" />
 			</output>
 		</layer>
-		<layer id="438" name="__module.model.gpt_neox.layers.1.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
+		<layer id="426" name="Constant_38785" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="2" precision="I64" />
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="439" name="__module.model.gpt_neox.layers.1.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1">
+		<layer id="427" name="Gather_38786" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
-				<port id="0" precision="I64" />
+				<port id="0" precision="I64">
+					<dim>4</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="1" precision="I64" names="529,530,534,535" />
-			</output>
-		</layer>
-		<layer id="440" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="441" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1">
+		<layer id="428" name="Add_38789" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
+				<port id="2" precision="I64" names="564">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="442" name="Constant_269562" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="429" name="Constant_1780" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="443" name="ScatterUpdate_269565" type="ScatterUpdate" version="opset3">
+		<layer id="430" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Slice_14" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -6671,1089 +6253,1076 @@
 				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
 				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="444" name="Constant_269568" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="445" name="Constant_269571" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="5" precision="FP32" names="568">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="446" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="431" name="__module.model.gpt_neox.layers.1.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+			<data causal="false" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="3" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="536,x2.7">
+				<port id="4" precision="FP32" names="569,attn_output.7">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="447" name="Constant_299087_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
+		<layer id="432" name="__module.model.gpt_neox.layers.1.attention/aten::transpose/Constant" type="Const" version="opset1">
+			<data element_type="i32" shape="4" offset="201116200" size="16" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="I32">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="448" name="Constant_299087" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="433" name="__module.model.gpt_neox.layers.1.attention/aten::transpose/Transpose" type="Transpose" version="opset1">
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I32">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="570">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="449" name="__module.model.gpt_neox.layers.1.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="434" name="Constant_40985" type="Const" version="opset1">
+			<data element_type="i64" shape="3" offset="201116216" size="24" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>3</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="435" name="__module.model.gpt_neox.layers.1.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
+			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
+					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>3</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="573">
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="FP32">
+			</output>
+		</layer>
+		<layer id="436" name="self.model.gpt_neox.layers.1.attention.dense.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 2560" offset="279936080" size="6553600" />
+			<output>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="437" name="Convert_75187" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="438" name="self.model.gpt_neox.layers.1.attention.dense.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="286489680" size="2560" />
+			<output>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="439" name="Convert_75190" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
 					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="440" name="self.model.gpt_neox.layers.1.attention.dense.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="537">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="450" name="Constant_269616" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="441" name="self.model.gpt_neox.layers.1.attention.dense.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="286492240" size="5120" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="451" name="Constant_269621" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="442" name="self.model.gpt_neox.layers.1.attention.dense.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="443" name="self.model.gpt_neox.layers.1.attention.dense.weight/fq_weights_1/convert" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="452" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="444" name="__module.model.gpt_neox.layers.1.attention.dense/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="531">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="453" name="__module.model.gpt_neox.layers.1.attention/aten::cat/Concat_2" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="445" name="Constant_37303" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="286497360" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="446" name="__module.model.gpt_neox.layers.1.attention.dense/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="539">
+				<port id="2" precision="FP32" names="576,578,input.5">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="454" name="__module.model.gpt_neox.layers.1.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
+		<layer id="447" name="__module.model.gpt_neox.layers.1/aten::add/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
-					<dim>1</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="540">
+				<port id="2" precision="FP32" names="596">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="455" name="__module.model.gpt_neox.layers.1.attention/aten::add/Add_1" type="Add" version="opset1">
+		<layer id="448" name="__module.model.gpt_neox.layers.1/aten::add/Add_1" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="541">
+				<port id="2" precision="FP32" names="597,599,hidden_states.11">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="456" name="Constant_269666" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="457" name="Constant_269669" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="458" name="Constant_269672" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="449" name="__module.model.gpt_neox.layers.2.input_layernorm/aten::layer_norm/Multiply" type="Const" version="opset1">
+			<data element_type="i32" shape="1" offset="128867668" size="4" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="I32">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="459" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="450" name="__module.model.gpt_neox.layers.2.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
+			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="I32">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="489">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>60</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="460" name="__module.model.gpt_neox.layers.1.attention/aten::cat/Concat_3" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="451" name="Constant_37304" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="286507600" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="452" name="__module.model.gpt_neox.layers.2.post_attention_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="545">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="461" name="__module.model.gpt_neox.layers.1.attention/aten::to/Convert_1" type="Concat" version="opset1">
-			<data axis="-2" />
+		<layer id="453" name="Constant_37305" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="286517840" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="454" name="__module.model.gpt_neox.layers.2.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="present.1.key">
+				<port id="2" precision="FP32" names="697">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="462" name="Constant_337468_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
+		<layer id="455" name="self.model.gpt_neox.layers.2.mlp.dense_h_to_4h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 2560" offset="286528080" size="26214400" />
 			<output>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="463" name="Constant_337468" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="456" name="Convert_74224" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" />
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="464" name="Constant_337461" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
+		<layer id="457" name="self.model.gpt_neox.layers.2.mlp.dense_h_to_4h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 1" offset="312742480" size="10240" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="U8">
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="465" name="Constant_337463" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64">
+		<layer id="458" name="Convert_74227" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="466" name="Constant_337465" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
+			</input>
 			<output>
-				<port id="0" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="467" name="Concat_337466" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="459" name="self.model.gpt_neox.layers.2.mlp.dense_h_to_4h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="468" name="Broadcast_337469" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
+		<layer id="460" name="self.model.gpt_neox.layers.2.mlp.dense_h_to_4h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="10240, 1" offset="312752720" size="20480" />
 			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="469" name="ReadValue_330431" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.1.valuepresent.1.value" variable_type="f32" variable_shape="?,32,?,80" />
+		<layer id="461" name="self.model.gpt_neox.layers.2.mlp.dense_h_to_4h.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="past_key_values.1.value">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="470" name="Constant_323394" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
+		<layer id="462" name="__module.model.gpt_neox.layers.2.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<rt_info>
+				<attribute name="decompression" version="0" />
+			</rt_info>
+			<input>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64" />
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="471" name="Gather_323395" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="463" name="__module.model.gpt_neox.layers.2.mlp.dense_h_to_4h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="3" precision="FP32">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="472" name="Constant_269678" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="464" name="self.model.gpt_neox.layers.2.mlp.dense_h_to_4h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="10240" offset="312773200" size="20480" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.2.mlp.dense_h_to_4h.bias">
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="473" name="Constant_269681" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
+		<layer id="465" name="__module.model.gpt_neox.layers.2.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<rt_info>
-				<attribute name="precise" version="0" />
+				<attribute name="decompression" version="0" />
 			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+			<input>
+				<port id="0" precision="BF16">
+					<dim>10240</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="474" name="Constant_269684" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="475" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="466" name="__module.model.gpt_neox.layers.2.mlp.dense_h_to_4h/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="483">
+				<port id="2" precision="FP32" names="702">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="476" name="Constant_832" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="484">
-					<dim>4</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="477" name="__module.model.gpt_neox.layers.1.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
+		<layer id="467" name="__module.model.gpt_neox.layers.2.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
+			<data approximation_mode="ERF" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="485">
+				<port id="1" precision="FP32" names="703">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="478" name="__module.model.gpt_neox.layers.1.attention/aten::cat/Concat_5" type="Concat" version="opset1">
-			<data axis="-2" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+		<layer id="468" name="self.model.gpt_neox.layers.2.mlp.dense_4h_to_h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 10240" offset="312793680" size="26214400" />
+			<output>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+			</output>
+		</layer>
+		<layer id="469" name="Convert_74234" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="present.1.value">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="479" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Slice_11" type="Const" version="opset1">
-			<data element_type="boolean" shape="1, 1, 2048, 2048" offset="201244276" size="4194304" />
+		<layer id="470" name="self.model.gpt_neox.layers.2.mlp.dense_4h_to_h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="339008080" size="2560" />
 			<output>
-				<port id="0" precision="BOOL" names="562">
-					<dim>1</dim>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
 					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="480" name="Constant_269688" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+		<layer id="471" name="Convert_74237" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="481" name="Constant_269687" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162298" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="482" name="__module.model.gpt_neox.layers.1.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="472" name="self.model.gpt_neox.layers.2.mlp.dense_4h_to_h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="483" name="__module.model.gpt_neox.layers.1.attention/aten::size/Constant_8" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="473" name="self.model.gpt_neox.layers.2.mlp.dense_4h_to_h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="339010640" size="5120" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="484" name="__module.model.gpt_neox.layers.1.attention/aten::size/Gather_8" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="557" />
 			</output>
 		</layer>
-		<layer id="485" name="__module.model.gpt_neox.layers.1.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="474" name="self.model.gpt_neox.layers.2.mlp.dense_4h_to_h.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="486" name="__module.model.gpt_neox.layers.1.attention/aten::size/Constant_9" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
+		<layer id="475" name="__module.model.gpt_neox.layers.2.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<rt_info>
-				<attribute name="precise" version="0" />
+				<attribute name="decompression" version="0" />
 			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="487" name="__module.model.gpt_neox.layers.1.attention/aten::size/Gather_9" type="Gather" version="opset8">
-			<data batch_dims="0" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="555" />
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="488" name="__module.model.gpt_neox.layers.1.attention/aten::sub/Subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="476" name="__module.model.gpt_neox.layers.2.mlp.dense_4h_to_h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64" names="559,560" />
-			</output>
-		</layer>
-		<layer id="489" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Constant_4" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="490" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="477" name="self.model.gpt_neox.layers.2.mlp.dense_4h_to_h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="2560" offset="339015760" size="5120" />
 			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.2.mlp.dense_4h_to_h.bias">
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="491" name="Constant_269686" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
+		<layer id="478" name="__module.model.gpt_neox.layers.2.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<rt_info>
-				<attribute name="precise" version="0" />
+				<attribute name="decompression" version="0" />
 			</rt_info>
+			<input>
+				<port id="0" precision="BF16">
+					<dim>2560</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="492" name="ScatterUpdate_269689" type="ScatterUpdate" version="opset3">
+		<layer id="479" name="__module.model.gpt_neox.layers.2.mlp.dense_4h_to_h/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="493" name="Constant_269690" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="2" precision="FP32" names="706,input.11">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="494" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="480" name="Constant_37306" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="339020880" size="10240" />
 			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="FP32">
+					<dim>1</dim>
 					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="495" name="ScatterUpdate_269691" type="ScatterUpdate" version="opset3">
+		<layer id="481" name="__module.model.gpt_neox.layers.2.input_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="2" precision="I64">
+				<port id="1" precision="FP32">
 					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
 					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="496" name="Constant_269694" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438612" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="482" name="Constant_37307" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="339031120" size="10240" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="497" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="483" name="__module.model.gpt_neox.layers.2.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="BOOL">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>3</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="BOOL" names="563">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="609,hidden_states.9">
 					<dim>-1</dim>
-					<dim>2048</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="498" name="Constant_269767" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="484" name="self.model.gpt_neox.layers.2.attention.query_key_value.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 2560" offset="339041360" size="19660800" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="499" name="Constant_269768" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="485" name="Convert_74846" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="500" name="Constant_269764" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="486" name="self.model.gpt_neox.layers.2.attention.query_key_value.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 1" offset="358702160" size="7680" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="U8">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="501" name="Constant_269763" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="487" name="Convert_74849" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>1</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I32">
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="502" name="ScatterUpdate_269769" type="ScatterUpdate" version="opset3">
+		<layer id="488" name="self.model.gpt_neox.layers.2.attention.query_key_value.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="503" name="Constant_269772" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="489" name="self.model.gpt_neox.layers.2.attention.query_key_value.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="7680, 1" offset="358709840" size="15360" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="504" name="__module.model.gpt_neox.layers.1.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="490" name="self.model.gpt_neox.layers.2.attention.query_key_value.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="BOOL" names="564,565">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
+				<port id="2" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="505" name="__module.model.gpt_neox.layers.1.attention/aten::where/Convert_1_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="506" name="__module.model.gpt_neox.layers.1.attention/aten::where/Convert_1" type="Convert" version="opset1">
+		<layer id="491" name="self.model.gpt_neox.layers.2.attention.query_key_value.weight/fq_weights_1/convert" type="Convert" version="opset1">
 			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
 			<input>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="507" name="__module.model.gpt_neox.layers.1.attention/aten::full/Convert" type="Const" version="opset1">
-			<data element_type="f32" shape="" offset="205438636" size="4" />
-			<output>
-				<port id="0" precision="FP32" names="552" />
+				<port id="1" precision="FP32">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="508" name="__module.model.gpt_neox.layers.1.attention/aten::where/Select" type="Select" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="492" name="__module.model.gpt_neox.layers.2.attention.query_key_value/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="FP32" />
-				<port id="2" precision="FP32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="566,causal_mask.5">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="509" name="Add_290344" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
+		<layer id="493" name="Constant_37308" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 7680" offset="358725200" size="30720" />
+			<output>
 				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
+					<dim>7680</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="494" name="__module.model.gpt_neox.layers.2.attention.query_key_value/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>7680</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>-1</dim>
+					<dim>7680</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="616,qkv.9">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="510" name="ShapeOf_290345" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="495" name="__module.model.gpt_neox.layers.2.attention/prim::ListConstruct" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201115992" size="32" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="496" name="__module.model.gpt_neox.layers.2.attention/aten::view/Reshape" type="Reshape" version="opset1">
+			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>7680</dim>
 				</port>
-			</input>
-			<output>
 				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="620,qkv.11">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>240</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="511" name="Constant_254344" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="497" name="Constant_1976" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="512" name="__module.model.gpt_neox.layers.1.attention/aten::size/Constant_10" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="498" name="Constant_1978" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="513" name="__module.model.gpt_neox.layers.1.attention/aten::size/Gather_10" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I32" />
-			</input>
+			</output>
+		</layer>
+		<layer id="499" name="Constant_1980" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="3" precision="I64" names="550">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="514" name="__module.model.gpt_neox.layers.1.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="500" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Slice" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -7764,1269 +7333,1169 @@
 				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64" names="567">
+				<port id="5" precision="FP32" names="621">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="501" name="Constant_1989" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
+			<output>
+				<port id="0" precision="I64" names="622">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="515" name="__module.model.gpt_neox.layers.1.attention/aten::expand/Abs" type="Abs" version="opset1">
+		<layer id="502" name="__module.model.gpt_neox.layers.2.attention/aten::permute/Transpose" type="Transpose" version="opset1">
 			<input>
-				<port id="0" precision="I64">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="623,query.5">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="516" name="Maximum_290346" type="Maximum" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
+		<layer id="503" name="Constant_2030" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+			</output>
+		</layer>
+		<layer id="504" name="Constant_2032" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="505" name="Constant_2034" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="517" name="__module.model.gpt_neox.layers.1.attention/aten::add/Add_2" type="Broadcast" version="opset3">
-			<data mode="numpy" />
+		<layer id="506" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Slice_1" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="569">
-					<dim>-1</dim>
-					<dim>-1</dim>
+				<port id="5" precision="FP32" names="630,q.5">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="518" name="__module.model.gpt_neox.layers.1.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
-			<data causal="false" />
+		<layer id="507" name="__module.model.gpt_neox.layers.2.attention/aten::mul/Multiply" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
 				</port>
-				<port id="2" precision="FP32">
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="636">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
 				</port>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
+			</output>
+		</layer>
+		<layer id="508" name="Constant_2188" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="509" name="Constant_2190" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
-				<port id="4" precision="FP32" names="570,571,sdpa_result.3,tensor.5">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="519" name="Constant_1273" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="510" name="Constant_2192" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="I64" names="572">
-					<dim>4</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="520" name="__module.model.gpt_neox.layers.1.attention/aten::permute/Transpose_3" type="Transpose" version="opset1">
+		<layer id="511" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Slice_2" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="573">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="521" name="__module.model.gpt_neox.layers.1.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438650" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="2" precision="I64">
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="522" name="__module.model.gpt_neox.layers.1.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
-			<data special_zero="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
+				<port id="3" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>3</dim>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="578">
+				<port id="5" precision="FP32" names="646,x2.9">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="523" name="self.model.gpt_neox.layers.1.attention.dense.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 2560" offset="284217554" size="6553600" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="524" name="Convert_21474" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="525" name="self.model.gpt_neox.layers.1.attention.dense.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="290771154" size="2560" />
+		<layer id="512" name="Constant_37309" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
 					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="526" name="Convert_21477" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
 					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="527" name="self.model.gpt_neox.layers.1.attention.dense.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="513" name="__module.model.gpt_neox.layers.2.attention/aten::neg/Multiply" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32" names="647">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="528" name="self.model.gpt_neox.layers.1.attention.dense.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="290773714" size="5120" />
+		<layer id="514" name="Constant_2085" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="529" name="self.model.gpt_neox.layers.1.attention.dense.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
+		<layer id="515" name="Constant_2129" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="530" name="self.model.gpt_neox.layers.1.attention.dense.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="516" name="Constant_2131" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="531" name="__module.model.gpt_neox.layers.1.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
+		<layer id="517" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Slice_3" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="532" name="Constant_299088_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="290778834" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="533" name="Constant_299088" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
+				<port id="3" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+				<port id="5" precision="FP32" names="641">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="534" name="__module.model.gpt_neox.layers.1.attention.dense/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="518" name="__module.model.gpt_neox.layers.2.attention/aten::cat/Concat" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="581,583,input.7">
+				<port id="2" precision="FP32" names="649">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="535" name="__module.model.gpt_neox.layers.1/aten::add/Add" type="Add" version="opset1">
+		<layer id="519" name="__module.model.gpt_neox.layers.2.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="601">
+				<port id="2" precision="FP32" names="650">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="536" name="__module.model.gpt_neox.layers.1/aten::add/Add_1" type="Add" version="opset1">
+		<layer id="520" name="__module.model.gpt_neox.layers.2.attention/aten::add/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="602,604,hidden_states.5">
+				<port id="2" precision="FP32" names="651">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="537" name="__module.model.gpt_neox.layers.2.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867670" size="4" />
+		<layer id="521" name="Constant_2040" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="538" name="__module.model.gpt_neox.layers.2.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
-			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
+		<layer id="522" name="Constant_2042" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="539" name="Constant_299089_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="290783954" size="5120" />
+		<layer id="523" name="Constant_2044" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="540" name="Constant_299089" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="524" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Slice_4" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
+				<port id="3" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="5" precision="FP32" names="631">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="541" name="__module.model.gpt_neox.layers.2.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="525" name="__module.model.gpt_neox.layers.2.attention/aten::cat/Concat_1" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="669">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="542" name="Constant_299090_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="290789074" size="5120" />
+		<layer id="526" name="Constant_52599" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="FP32" />
+			</output>
+		</layer>
+		<layer id="527" name="Constant_52592" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="543" name="Constant_299090" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+		<layer id="528" name="Constant_52594" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="529" name="Constant_52596" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="544" name="__module.model.gpt_neox.layers.2.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="530" name="Concat_52597" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="FP32">
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="739">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
+				<port id="4" precision="I64">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="545" name="self.model.gpt_neox.layers.2.mlp.dense_h_to_4h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 2560" offset="290794194" size="26214400" />
+		<layer id="531" name="Broadcast_52600" type="Broadcast" version="opset3">
+			<data mode="numpy" />
+			<input>
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
+					<dim>4</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="546" name="Convert_21524" type="Convert" version="opset1">
-			<data destination_type="f16" />
+		<layer id="532" name="ReadValue_50231" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.2.keypresent.2.key" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="1" precision="FP32" names="past_key_values.2.key">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="547" name="self.model.gpt_neox.layers.2.mlp.dense_h_to_4h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 1" offset="317008594" size="10240" />
+		<layer id="533" name="Constant_47790" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="548" name="Convert_21527" type="Convert" version="opset1">
-			<data destination_type="f16" />
+		<layer id="534" name="Gather_47791" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>1</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I32">
+					<dim>-1</dim>
 				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
+				<port id="3" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="549" name="self.model.gpt_neox.layers.2.mlp.dense_h_to_4h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
+		<layer id="535" name="Constant_1994" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="536" name="Constant_1996" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
-				<port id="2" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="550" name="self.model.gpt_neox.layers.2.mlp.dense_h_to_4h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="10240, 1" offset="317018834" size="20480" />
+		<layer id="537" name="Constant_1998" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="551" name="self.model.gpt_neox.layers.2.mlp.dense_h_to_4h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="538" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Slice_5" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="5" precision="FP32" names="624">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="552" name="self.model.gpt_neox.layers.2.mlp.dense_h_to_4h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="539" name="Constant_2007" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64" names="625">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="553" name="__module.model.gpt_neox.layers.2.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
+		<layer id="540" name="__module.model.gpt_neox.layers.2.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="626,key.9">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="554" name="Constant_299091_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 10240" offset="317039314" size="20480" />
+		<layer id="541" name="Constant_2050" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="555" name="Constant_299091" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+		<layer id="542" name="Constant_2052" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>10240</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="543" name="Constant_2054" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="556" name="__module.model.gpt_neox.layers.2.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="544" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Slice_6" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="744">
+				<port id="5" precision="FP32" names="632,k.5">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="557" name="__module.model.gpt_neox.layers.2.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
-			<data approximation_mode="ERF" />
+		<layer id="545" name="__module.model.gpt_neox.layers.2.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="745">
+				<port id="2" precision="FP32" names="652">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="558" name="self.model.gpt_neox.layers.2.mlp.dense_4h_to_h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 10240" offset="317059794" size="26214400" />
+		<layer id="546" name="Constant_2331" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="559" name="Convert_21534" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="560" name="self.model.gpt_neox.layers.2.mlp.dense_4h_to_h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="343274194" size="2560" />
+		<layer id="547" name="Constant_2333" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="561" name="Convert_21537" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
+		<layer id="548" name="Constant_2335" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="562" name="self.model.gpt_neox.layers.2.mlp.dense_4h_to_h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="549" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Slice_7" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="563" name="self.model.gpt_neox.layers.2.mlp.dense_4h_to_h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="343276754" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
+				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="564" name="self.model.gpt_neox.layers.2.mlp.dense_4h_to_h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="3" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="5" precision="FP32" names="662,x2.11">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="565" name="self.model.gpt_neox.layers.2.mlp.dense_4h_to_h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
+		<layer id="550" name="Constant_37310" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="566" name="__module.model.gpt_neox.layers.2.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
+		<layer id="551" name="__module.model.gpt_neox.layers.2.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>10</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="663">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="567" name="Constant_299092_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="343281874" size="5120" />
+		<layer id="552" name="Constant_2228" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="568" name="Constant_299092" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+		<layer id="553" name="Constant_2272" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="554" name="Constant_2274" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="569" name="__module.model.gpt_neox.layers.2.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="555" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Slice_8" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="1" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="748,input.13">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="570" name="Constant_299093_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="343286994" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="2" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="571" name="Constant_299093" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
+				<port id="3" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+				<port id="5" precision="FP32" names="657">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="572" name="__module.model.gpt_neox.layers.2.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="556" name="__module.model.gpt_neox.layers.2.attention/aten::cat/Concat_2" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="665">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="573" name="Constant_299094_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="343292114" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="574" name="Constant_299094" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="557" name="__module.model.gpt_neox.layers.2.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32" names="666">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="575" name="__module.model.gpt_neox.layers.2.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+		<layer id="558" name="__module.model.gpt_neox.layers.2.attention/aten::add/Add_1" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="614">
+				<port id="2" precision="FP32" names="667">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="576" name="self.model.gpt_neox.layers.2.attention.query_key_value.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 2560" offset="343297234" size="19660800" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="577" name="Convert_21504" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="559" name="Constant_2060" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="578" name="self.model.gpt_neox.layers.2.attention.query_key_value.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 1" offset="362958034" size="7680" />
+		<layer id="560" name="Constant_2062" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="579" name="Convert_21507" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</input>
+		<layer id="561" name="Constant_2064" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="580" name="self.model.gpt_neox.layers.2.attention.query_key_value.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="562" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Slice_9" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
+				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="581" name="self.model.gpt_neox.layers.2.attention.query_key_value.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="7680, 1" offset="362965714" size="15360" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
+				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="582" name="self.model.gpt_neox.layers.2.attention.query_key_value.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="3" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="5" precision="FP32" names="633">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="583" name="self.model.gpt_neox.layers.2.attention.query_key_value.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="563" name="__module.model.gpt_neox.layers.2.attention/aten::cat/Concat_3" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32" names="671">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="584" name="__module.model.gpt_neox.layers.2.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
+		<layer id="564" name="__module.model.gpt_neox.layers.2.attention/aten::cat/Concat_4" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>7680</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="present.2.key">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="585" name="Constant_299095_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 7680" offset="362981074" size="15360" />
+		<layer id="565" name="Constant_52611" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32" />
+			</output>
+		</layer>
+		<layer id="566" name="Constant_52604" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="567" name="Constant_52606" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="568" name="Constant_52608" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="586" name="Constant_299095" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="569" name="Concat_52609" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
 					<dim>1</dim>
-					<dim>7680</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
+				<port id="4" precision="I64">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="587" name="__module.model.gpt_neox.layers.2.attention.query_key_value/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="570" name="Broadcast_52612" type="Broadcast" version="opset3">
+			<data mode="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
+					<dim>4</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
+			</output>
+		</layer>
+		<layer id="571" name="ReadValue_50233" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.2.valuepresent.2.value" variable_type="f32" variable_shape="?,32,?,80" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="621,qkv.9">
+				<port id="1" precision="FP32" names="past_key_values.2.value">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="588" name="__module.model.gpt_neox.layers.2.attention/prim::ListConstruct/Concat" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080154" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="572" name="Constant_47793" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="589" name="__module.model.gpt_neox.layers.2.attention/aten::view/Reshape" type="Reshape" version="opset1">
-			<data special_zero="true" />
+		<layer id="573" name="Gather_47794" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="I32">
+					<dim>-1</dim>
 				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="625,qkv.11">
-					<dim>-1</dim>
+				<port id="3" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>240</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="590" name="Constant_269779" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="574" name="Constant_2012" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="591" name="Constant_269782" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="575" name="Constant_2014" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="592" name="Constant_269785" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="576" name="Constant_2016" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="593" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="577" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Slice_10" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -9035,17 +8504,20 @@
 					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="626">
+				<port id="5" precision="FP32" names="627">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -9053,15 +8525,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="594" name="Constant_1385" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="578" name="Constant_2025" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64" names="627">
+				<port id="0" precision="I64" names="628">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="595" name="__module.model.gpt_neox.layers.2.attention/aten::permute/Transpose" type="Transpose" version="opset1">
+		<layer id="579" name="__module.model.gpt_neox.layers.2.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -9074,7 +8546,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="628,query.13">
+				<port id="2" precision="FP32" names="629">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -9082,41 +8554,41 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="596" name="Constant_269791" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+		<layer id="580" name="__module.model.gpt_neox.layers.2.attention/aten::cat/Concat_5" type="Concat" version="opset1">
+			<data axis="-2" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="597" name="Constant_269794" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="present.2.value">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="598" name="Constant_269797" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="581" name="Constant_2407" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="599" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="582" name="ShapeOf_38854" type="ShapeOf" version="opset3">
+			<data output_type="i64" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -9124,146 +8596,128 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
+			</input>
+			<output>
 				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="583" name="Constant_38855" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116168" size="8" />
 			<output>
-				<port id="4" precision="FP32" names="635,q.5">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="600" name="self.model.gpt_neox.layers.2.attention.rotary_emb.cos_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201080346" size="81920" />
+		<layer id="584" name="Constant_38856" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.2.attention.rotary_emb.cos_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="601" name="self.model.gpt_neox.layers.2.attention.rotary_emb.cos_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="585" name="Gather_38857" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="I64">
+					<dim>4</dim>
 				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="3" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="602" name="__module.model.gpt_neox.layers.2.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
+		<layer id="586" name="Add_38860" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="603" name="Constant_269803" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="I64">
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="604" name="Constant_269806" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="I64" names="676">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="605" name="Constant_269809" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="587" name="Constant_2410" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="606" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="588" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Slice_14" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="629">
+				<port id="5" precision="FP32" names="680">
+					<dim>-1</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="607" name="Constant_1399" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="630">
-					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="608" name="__module.model.gpt_neox.layers.2.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
+		<layer id="589" name="__module.model.gpt_neox.layers.2.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+			<data causal="false" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="3" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="631,key.13">
+				<port id="4" precision="FP32" names="681,attn_output.13">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -9271,8 +8725,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="609" name="__module.model.gpt_neox.layers.2.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="590" name="__module.model.gpt_neox.layers.2.attention/aten::transpose/Constant" type="Const" version="opset1">
+			<data element_type="i32" shape="4" offset="201116200" size="16" />
+			<output>
+				<port id="0" precision="I32">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="591" name="__module.model.gpt_neox.layers.2.attention/aten::transpose/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -9280,897 +8741,1052 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
+				<port id="1" precision="I32">
+					<dim>4</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="682">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="610" name="__module.model.gpt_neox.layers.2.attention/aten::size/Constant_2" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="592" name="Constant_40986" type="Const" version="opset1">
+			<data element_type="i64" shape="3" offset="201116216" size="24" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="I64">
+					<dim>3</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="611" name="__module.model.gpt_neox.layers.2.attention/aten::size/Gather_2" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="593" name="__module.model.gpt_neox.layers.2.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
+			<data special_zero="true" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>3</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="639" />
+				<port id="2" precision="FP32" names="685">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="612" name="Constant_337481_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
+		<layer id="594" name="self.model.gpt_neox.layers.2.attention.dense.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 2560" offset="358755920" size="6553600" />
 			<output>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="613" name="Constant_337481" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="595" name="Convert_75198" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" />
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="614" name="Constant_337474" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
+		<layer id="596" name="self.model.gpt_neox.layers.2.attention.dense.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="365309520" size="2560" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="U8">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="615" name="Constant_337476" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64">
+		<layer id="597" name="Convert_75201" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="616" name="Constant_337478" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
+			</input>
 			<output>
-				<port id="0" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="617" name="Concat_337479" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="598" name="self.model.gpt_neox.layers.2.attention.dense.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="599" name="self.model.gpt_neox.layers.2.attention.dense.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="365312080" size="5120" />
+			<output>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I64">
+			</output>
+		</layer>
+		<layer id="600" name="self.model.gpt_neox.layers.2.attention.dense.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="618" name="Broadcast_337482" type="Broadcast" version="opset3">
-			<data mode="numpy" />
+		<layer id="601" name="self.model.gpt_neox.layers.2.attention.dense.weight/fq_weights_1/convert" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="619" name="ReadValue_330473" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.2.keypresent.2.key" variable_type="f32" variable_shape="?,32,?,80" />
+		<layer id="602" name="__module.model.gpt_neox.layers.2.attention.dense/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="past_key_values.2.key">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="620" name="Constant_323397" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
+		<layer id="603" name="Constant_37311" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="365317200" size="10240" />
 			<output>
-				<port id="0" precision="I64" />
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="621" name="Gather_323398" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="604" name="__module.model.gpt_neox.layers.2.attention.dense/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="3" precision="FP32">
+				<port id="2" precision="FP32" names="688,690,input.9">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="622" name="__module.model.gpt_neox.layers.2.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="605" name="__module.model.gpt_neox.layers.2/aten::add/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="623" name="__module.model.gpt_neox.layers.2.attention/aten::size/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="624" name="__module.model.gpt_neox.layers.2.attention/aten::size/Gather_3" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="641" />
-			</output>
-		</layer>
-		<layer id="625" name="__module.model.gpt_neox.layers.2.attention/aten::add_/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" names="640,644,645,seq_len.9" />
-			</output>
-		</layer>
-		<layer id="626" name="__module.model.gpt_neox.layers.2.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="627" name="__module.model.gpt_neox.layers.2.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
 			</input>
 			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="628" name="__module.model.gpt_neox.layers.2.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="708">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="629" name="__module.model.gpt_neox.layers.2.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="606" name="__module.model.gpt_neox.layers.2/aten::add/Add_1" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="648,651,cos.9">
+				<port id="2" precision="FP32" names="709,711,hidden_states.15">
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="630" name="Constant_38209" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
+		<layer id="607" name="__module.model.gpt_neox.layers.3.input_layernorm/aten::layer_norm/Multiply" type="Const" version="opset1">
+			<data element_type="i32" shape="1" offset="128867668" size="4" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="I32">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="631" name="__module.model.gpt_neox.layers.2.attention/aten::index/Gather" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="608" name="__module.model.gpt_neox.layers.3.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
+			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
+					<dim>1</dim>
 				</port>
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="654">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="632" name="__module.model.gpt_neox.layers.2.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64" />
-			</input>
+		<layer id="609" name="Constant_37312" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="365327440" size="10240" />
 			<output>
-				<port id="2" precision="FP32" names="655,cos.11">
-					<dim>-1</dim>
+				<port id="0" precision="FP32">
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="633" name="__module.model.gpt_neox.layers.2.attention/aten::mul/Multiply" type="Multiply" version="opset1">
+		<layer id="610" name="__module.model.gpt_neox.layers.3.post_attention_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="659">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="634" name="Constant_269814" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="635" name="Constant_269813" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="611" name="Constant_37313" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="365337680" size="10240" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="FP32">
+					<dim>1</dim>
 					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="636" name="__module.model.gpt_neox.layers.2.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="612" name="__module.model.gpt_neox.layers.3.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="809">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="637" name="__module.model.gpt_neox.layers.2.attention/aten::size/Constant_5" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="613" name="self.model.gpt_neox.layers.3.mlp.dense_h_to_4h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 2560" offset="365347920" size="26214400" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="638" name="__module.model.gpt_neox.layers.2.attention/aten::size/Gather_5" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="614" name="Convert_74244" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="660,665" />
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="639" name="__module.model.gpt_neox.layers.2.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
+		<layer id="615" name="self.model.gpt_neox.layers.3.mlp.dense_h_to_4h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 1" offset="391562320" size="10240" />
 			<output>
-				<port id="2" precision="I64" />
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="640" name="__module.model.gpt_neox.layers.2.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1">
+		<layer id="616" name="Convert_74247" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64" />
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64" names="662,663,667,668" />
-			</output>
-		</layer>
-		<layer id="641" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Constant_1" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="642" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1">
+		<layer id="617" name="self.model.gpt_neox.layers.3.mlp.dense_h_to_4h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="643" name="Constant_269812" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="618" name="self.model.gpt_neox.layers.3.mlp.dense_h_to_4h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="10240, 1" offset="391572560" size="20480" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="644" name="ScatterUpdate_269815" type="ScatterUpdate" version="opset3">
+		<layer id="619" name="self.model.gpt_neox.layers.3.mlp.dense_h_to_4h.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="645" name="Constant_269818" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
+		<layer id="620" name="__module.model.gpt_neox.layers.3.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<rt_info>
-				<attribute name="precise" version="0" />
+				<attribute name="decompression" version="0" />
 			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+			<input>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="646" name="Constant_269821" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="647" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="621" name="__module.model.gpt_neox.layers.3.mlp.dense_h_to_4h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+			</input>
+			<output>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+			</output>
+		</layer>
+		<layer id="622" name="self.model.gpt_neox.layers.3.mlp.dense_h_to_4h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="10240" offset="391593040" size="20480" />
+			<output>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.3.mlp.dense_h_to_4h.bias">
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="623" name="__module.model.gpt_neox.layers.3.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<rt_info>
+				<attribute name="decompression" version="0" />
+			</rt_info>
+			<input>
+				<port id="0" precision="BF16">
+					<dim>10240</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="624" name="__module.model.gpt_neox.layers.3.mlp.dense_h_to_4h/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="669,x2.9">
+				<port id="2" precision="FP32" names="814">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="625" name="__module.model.gpt_neox.layers.3.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
+			<data approximation_mode="ERF" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP32" names="815">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="648" name="Constant_299096_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
+		<layer id="626" name="self.model.gpt_neox.layers.3.mlp.dense_4h_to_h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 10240" offset="391613520" size="26214400" />
 			<output>
-				<port id="0" precision="FP16">
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="627" name="Convert_74254" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="628" name="self.model.gpt_neox.layers.3.mlp.dense_4h_to_h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="417827920" size="2560" />
+			<output>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="629" name="Convert_74257" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="630" name="self.model.gpt_neox.layers.3.mlp.dense_4h_to_h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="631" name="self.model.gpt_neox.layers.3.mlp.dense_4h_to_h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="417830480" size="5120" />
+			<output>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="632" name="self.model.gpt_neox.layers.3.mlp.dense_4h_to_h.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="649" name="Constant_299096" type="Convert" version="opset1">
+		<layer id="633" name="__module.model.gpt_neox.layers.3.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
 			</rt_info>
 			<input>
 				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="650" name="__module.model.gpt_neox.layers.2.attention/aten::neg/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="634" name="__module.model.gpt_neox.layers.3.mlp.dense_4h_to_h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>10240</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="670">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="651" name="Constant_269866" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="635" name="self.model.gpt_neox.layers.3.mlp.dense_4h_to_h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="2560" offset="417835600" size="5120" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.3.mlp.dense_4h_to_h.bias">
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="652" name="Constant_269871" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
+		<layer id="636" name="__module.model.gpt_neox.layers.3.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<rt_info>
-				<attribute name="precise" version="0" />
+				<attribute name="decompression" version="0" />
 			</rt_info>
+			<input>
+				<port id="0" precision="BF16">
+					<dim>2560</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="653" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="637" name="__module.model.gpt_neox.layers.3.mlp.dense_4h_to_h/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="664">
+				<port id="2" precision="FP32" names="818,input.15">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="654" name="__module.model.gpt_neox.layers.2.attention/aten::cat/Concat" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="638" name="Constant_37314" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="417840720" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="639" name="__module.model.gpt_neox.layers.3.input_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="640" name="Constant_37315" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="417850960" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="641" name="__module.model.gpt_neox.layers.3.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="672">
+				<port id="2" precision="FP32" names="721,hidden_states.13">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="655" name="self.model.gpt_neox.layers.2.attention.rotary_emb.sin_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201162356" size="81920" />
+		<layer id="642" name="self.model.gpt_neox.layers.3.attention.query_key_value.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 2560" offset="417861200" size="19660800" />
 			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.2.attention.rotary_emb.sin_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="656" name="self.model.gpt_neox.layers.2.attention.rotary_emb.sin_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="643" name="Convert_74857" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="657" name="__module.model.gpt_neox.layers.2.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="644" name="self.model.gpt_neox.layers.3.attention.query_key_value.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 1" offset="437522000" size="7680" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="U8">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="658" name="__module.model.gpt_neox.layers.2.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="645" name="Convert_74860" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>1</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="659" name="__module.model.gpt_neox.layers.2.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="646" name="self.model.gpt_neox.layers.3.attention.query_key_value.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="647" name="self.model.gpt_neox.layers.3.attention.query_key_value.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="7680, 1" offset="437529680" size="15360" />
+			<output>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I64">
+			</output>
+		</layer>
+		<layer id="648" name="self.model.gpt_neox.layers.3.attention.query_key_value.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="649,652,sin.9">
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="2" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="660" name="Constant_38212" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
+		<layer id="649" name="self.model.gpt_neox.layers.3.attention.query_key_value.weight/fq_weights_1/convert" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I32" />
+				<port id="1" precision="FP32">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="661" name="__module.model.gpt_neox.layers.2.attention/aten::index/Gather_1" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="650" name="__module.model.gpt_neox.layers.3.attention.query_key_value/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="657">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="662" name="__module.model.gpt_neox.layers.2.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64" />
-			</input>
+		<layer id="651" name="Constant_37316" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 7680" offset="437545040" size="30720" />
 			<output>
-				<port id="2" precision="FP32" names="658,sin.11">
-					<dim>-1</dim>
+				<port id="0" precision="FP32">
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>1</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="663" name="__module.model.gpt_neox.layers.2.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
+		<layer id="652" name="__module.model.gpt_neox.layers.3.attention.query_key_value/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>7680</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>1</dim>
+					<dim>7680</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="673">
+				<port id="2" precision="FP32" names="728,qkv.13">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="664" name="__module.model.gpt_neox.layers.2.attention/aten::add/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="653" name="__module.model.gpt_neox.layers.3.attention/prim::ListConstruct" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201115992" size="32" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="654" name="__module.model.gpt_neox.layers.3.attention/aten::view/Reshape" type="Reshape" version="opset1">
+			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>7680</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="674">
+				<port id="2" precision="FP32" names="732,qkv.15">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="665" name="Constant_269917" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="655" name="Constant_2606" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="666" name="Constant_269920" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="656" name="Constant_2608" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="667" name="Constant_269923" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="657" name="Constant_2610" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="668" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="658" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Slice" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="636">
+				<port id="5" precision="FP32" names="733">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>60</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="669" name="__module.model.gpt_neox.layers.2.attention/aten::cat/Concat_1" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="659" name="Constant_2619" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
+			<output>
+				<port id="0" precision="I64" names="734">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="660" name="__module.model.gpt_neox.layers.3.attention/aten::permute/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="692,702,query.15,query.17">
+				<port id="2" precision="FP32" names="735,query.7">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -10178,41 +9794,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="670" name="Constant_269929" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="661" name="Constant_2660" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="671" name="Constant_269932" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="662" name="Constant_2662" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="672" name="Constant_269935" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="663" name="Constant_2664" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="673" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="664" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Slice_1" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -10221,17 +9827,20 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="637,k.5">
+				<port id="5" precision="FP32" names="742,q.7">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -10239,7 +9848,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="674" name="__module.model.gpt_neox.layers.2.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
+		<layer id="665" name="__module.model.gpt_neox.layers.3.attention/aten::mul/Multiply" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -10256,7 +9865,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="675">
+				<port id="2" precision="FP32" names="748">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -10264,30 +9873,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="675" name="Constant_269939" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="666" name="Constant_2818" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="676" name="Constant_269938" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="667" name="Constant_2820" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="677" name="__module.model.gpt_neox.layers.2.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="668" name="Constant_2822" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="669" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Slice_2" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -10295,173 +9905,49 @@
 					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="5" precision="FP32" names="758,x2.13">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="678" name="__module.model.gpt_neox.layers.2.attention/aten::size/Constant_7" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="670" name="Constant_37317" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="679" name="__module.model.gpt_neox.layers.2.attention/aten::size/Gather_7" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="676,681" />
 			</output>
 		</layer>
-		<layer id="680" name="__module.model.gpt_neox.layers.2.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" />
-			</output>
-		</layer>
-		<layer id="681" name="__module.model.gpt_neox.layers.2.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-			</input>
-			<output>
-				<port id="1" precision="I64" names="678,679,683,684" />
-			</output>
-		</layer>
-		<layer id="682" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="683" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="684" name="Constant_269937" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="685" name="ScatterUpdate_269940" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="686" name="Constant_269943" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="687" name="Constant_269946" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="688" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="671" name="__module.model.gpt_neox.layers.3.attention/aten::neg/Multiply" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="685,x2.11">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
 					<dim>10</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="689" name="Constant_299097_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="690" name="Constant_299097" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
+				<port id="1" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>1</dim>
@@ -10469,63 +9955,39 @@
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="691" name="__module.model.gpt_neox.layers.2.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
+				<port id="2" precision="FP32" names="759">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
 					<dim>10</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="672" name="Constant_2715" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="2" precision="FP32" names="686">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="692" name="Constant_269991" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="673" name="Constant_2759" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="693" name="Constant_269996" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="674" name="Constant_2761" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="694" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="675" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Slice_3" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -10534,17 +9996,20 @@
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="680">
+				<port id="5" precision="FP32" names="753">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -10552,7 +10017,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="695" name="__module.model.gpt_neox.layers.2.attention/aten::cat/Concat_2" type="Concat" version="opset1">
+		<layer id="676" name="__module.model.gpt_neox.layers.3.attention/aten::cat/Concat" type="Concat" version="opset1">
 			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
@@ -10569,7 +10034,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="688">
+				<port id="2" precision="FP32" names="761">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -10577,7 +10042,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="696" name="__module.model.gpt_neox.layers.2.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
+		<layer id="677" name="__module.model.gpt_neox.layers.3.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -10594,7 +10059,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="689">
+				<port id="2" precision="FP32" names="762">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -10602,7 +10067,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="697" name="__module.model.gpt_neox.layers.2.attention/aten::add/Add_1" type="Add" version="opset1">
+		<layer id="678" name="__module.model.gpt_neox.layers.3.attention/aten::add/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -10619,7 +10084,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="690">
+				<port id="2" precision="FP32" names="763">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -10627,41 +10092,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="698" name="Constant_270041" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="679" name="Constant_2670" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="699" name="Constant_270044" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="680" name="Constant_2672" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="700" name="Constant_270047" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="681" name="Constant_2674" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="701" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="682" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Slice_4" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -10670,17 +10125,20 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="638">
+				<port id="5" precision="FP32" names="743">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -10688,7 +10146,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="702" name="__module.model.gpt_neox.layers.2.attention/aten::cat/Concat_3" type="Concat" version="opset1">
+		<layer id="683" name="__module.model.gpt_neox.layers.3.attention/aten::cat/Concat_1" type="Concat" version="opset1">
 			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
@@ -10705,32 +10163,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="694">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="703" name="__module.model.gpt_neox.layers.2.attention/aten::to/Convert_1" type="Concat" version="opset1">
-			<data axis="-2" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="present.2.key">
+				<port id="2" precision="FP32" names="781">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -10738,49 +10171,37 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="704" name="Constant_337494_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="705" name="Constant_337494" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
+		<layer id="684" name="Constant_52623" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="1" precision="FP32" />
+				<port id="0" precision="FP32" />
 			</output>
 		</layer>
-		<layer id="706" name="Constant_337487" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
+		<layer id="685" name="Constant_52616" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="707" name="Constant_337489" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
+		<layer id="686" name="Constant_52618" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="708" name="Constant_337491" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
+		<layer id="687" name="Constant_52620" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="709" name="Concat_337492" type="Concat" version="opset1">
+		<layer id="688" name="Concat_52621" type="Concat" version="opset1">
 			<data axis="0" />
 			<input>
 				<port id="0" precision="I64">
@@ -10802,7 +10223,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="710" name="Broadcast_337495" type="Broadcast" version="opset3">
+		<layer id="689" name="Broadcast_52624" type="Broadcast" version="opset3">
 			<data mode="numpy" />
 			<input>
 				<port id="0" precision="FP32" />
@@ -10819,8 +10240,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="711" name="ReadValue_330475" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.2.valuepresent.2.value" variable_type="f32" variable_shape="?,32,?,80" />
+		<layer id="690" name="ReadValue_50275" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.3.keypresent.3.key" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -10830,7 +10251,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="past_key_values.2.value">
+				<port id="1" precision="FP32" names="past_key_values.3.key">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -10838,13 +10259,13 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="712" name="Constant_323400" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
+		<layer id="691" name="Constant_47796" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="713" name="Gather_323401" type="Gather" version="opset8">
+		<layer id="692" name="Gather_47797" type="Gather" version="opset8">
 			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="FP32">
@@ -10867,41 +10288,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="714" name="Constant_270053" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="693" name="Constant_2624" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="715" name="Constant_270056" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="694" name="Constant_2626" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="716" name="Constant_270059" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="695" name="Constant_2628" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="717" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="696" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Slice_5" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -10910,17 +10321,20 @@
 					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="632">
+				<port id="5" precision="FP32" names="736">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -10928,15 +10342,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="718" name="Constant_1413" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="697" name="Constant_2637" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64" names="633">
+				<port id="0" precision="I64" names="737">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="719" name="__module.model.gpt_neox.layers.2.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
+		<layer id="698" name="__module.model.gpt_neox.layers.3.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -10949,32 +10363,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="634">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="720" name="__module.model.gpt_neox.layers.2.attention/aten::cat/Concat_5" type="Concat" version="opset1">
-			<data axis="-2" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="present.2.value">
+				<port id="2" precision="FP32" names="738,key.13">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -10982,41 +10371,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="721" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Slice_11" type="Const" version="opset1">
-			<data element_type="boolean" shape="1, 1, 2048, 2048" offset="201244276" size="4194304" />
+		<layer id="699" name="Constant_2680" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="BOOL" names="711">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="722" name="Constant_270063" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="700" name="Constant_2682" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>3</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="723" name="Constant_270062" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162298" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="701" name="Constant_2684" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="724" name="__module.model.gpt_neox.layers.2.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="702" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Slice_6" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -11024,118 +10403,84 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-			</input>
-			<output>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="725" name="__module.model.gpt_neox.layers.2.attention/aten::size/Constant_8" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="726" name="__module.model.gpt_neox.layers.2.attention/aten::size/Gather_8" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="706" />
+				<port id="5" precision="FP32" names="744,k.7">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="727" name="__module.model.gpt_neox.layers.2.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="703" name="__module.model.gpt_neox.layers.3.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="764">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="728" name="__module.model.gpt_neox.layers.2.attention/aten::size/Constant_9" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="704" name="Constant_2961" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="729" name="__module.model.gpt_neox.layers.2.attention/aten::size/Gather_9" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="704" />
-			</output>
-		</layer>
-		<layer id="730" name="__module.model.gpt_neox.layers.2.attention/aten::sub/Subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" names="708,709" />
-			</output>
-		</layer>
-		<layer id="731" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Constant_4" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
 			</output>
 		</layer>
-		<layer id="732" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="705" name="Constant_2963" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="733" name="Constant_270061" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="706" name="Constant_2965" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="734" name="ScatterUpdate_270064" type="ScatterUpdate" version="opset3">
+		<layer id="707" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Slice_7" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -11143,146 +10488,89 @@
 				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
 				<port id="4" precision="I64">
-					<dim>3</dim>
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="735" name="Constant_270065" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="5" precision="FP32" names="774,x2.15">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="736" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="708" name="Constant_37318" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="737" name="ScatterUpdate_270066" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
 					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
 					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="738" name="Constant_270069" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438612" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
 			</output>
 		</layer>
-		<layer id="739" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="709" name="__module.model.gpt_neox.layers.3.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="BOOL">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+				<port id="1" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>3</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="BOOL" names="712">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="740" name="Constant_270142" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="775">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="741" name="Constant_270143" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="710" name="Constant_2858" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="742" name="Constant_270139" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="711" name="Constant_2902" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="743" name="Constant_270138" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="712" name="Constant_2904" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="744" name="ScatterUpdate_270144" type="ScatterUpdate" version="opset3">
+		<layer id="713" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Slice_8" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -11290,184 +10578,350 @@
 				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
 				<port id="4" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="745" name="Constant_270147" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="5" precision="FP32" names="769">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="746" name="__module.model.gpt_neox.layers.2.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="714" name="__module.model.gpt_neox.layers.3.attention/aten::cat/Concat_2" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="BOOL" names="713,714">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="777">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="747" name="__module.model.gpt_neox.layers.2.attention/aten::where/Convert_1_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="748" name="__module.model.gpt_neox.layers.2.attention/aten::where/Convert_1" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="749" name="__module.model.gpt_neox.layers.2.attention/aten::full/Convert" type="Const" version="opset1">
-			<data element_type="f32" shape="" offset="205438636" size="4" />
-			<output>
-				<port id="0" precision="FP32" names="701" />
-			</output>
-		</layer>
-		<layer id="750" name="__module.model.gpt_neox.layers.2.attention/aten::where/Select" type="Select" version="opset1">
+		<layer id="715" name="__module.model.gpt_neox.layers.3.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="FP32" />
-				<port id="2" precision="FP32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="715,causal_mask.9">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="778">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="751" name="Add_290350" type="Add" version="opset1">
+		<layer id="716" name="__module.model.gpt_neox.layers.3.attention/aten::add/Add_1" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="779">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="717" name="Constant_2690" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="718" name="Constant_2692" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="719" name="Constant_2694" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="720" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Slice_9" type="Slice" version="opset8">
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="5" precision="FP32" names="745">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="721" name="__module.model.gpt_neox.layers.3.attention/aten::cat/Concat_3" type="Concat" version="opset1">
+			<data axis="-1" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="783">
 					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="752" name="ShapeOf_290351" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="722" name="__module.model.gpt_neox.layers.3.attention/aten::cat/Concat_4" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="present.3.key">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="753" name="Constant_254359" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="723" name="Constant_52635" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
+			<output>
+				<port id="0" precision="FP32" />
+			</output>
+		</layer>
+		<layer id="724" name="Constant_52628" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="754" name="__module.model.gpt_neox.layers.2.attention/aten::size/Constant_10" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="725" name="Constant_52630" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="755" name="__module.model.gpt_neox.layers.2.attention/aten::size/Gather_10" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="726" name="Constant_52632" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="727" name="Concat_52633" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I32" />
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="3" precision="I64" names="699">
-					<dim>1</dim>
+				<port id="4" precision="I64">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="756" name="__module.model.gpt_neox.layers.2.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="728" name="Broadcast_52636" type="Broadcast" version="opset3">
+			<data mode="numpy" />
+			<input>
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
+					<dim>4</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="729" name="ReadValue_50277" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.3.valuepresent.3.value" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP32" names="past_key_values.3.value">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="730" name="Constant_47799" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64" />
+			</output>
+		</layer>
+		<layer id="731" name="Gather_47800" type="Gather" version="opset8">
+			<data batch_dims="0" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I32">
+					<dim>-1</dim>
+				</port>
+				<port id="2" precision="I64" />
+			</input>
+			<output>
+				<port id="3" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="732" name="Constant_2642" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
+			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
+			</output>
+		</layer>
+		<layer id="733" name="Constant_2644" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="734" name="Constant_2646" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="735" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Slice_10" type="Slice" version="opset8">
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>240</dim>
+				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
@@ -11477,43 +10931,153 @@
 				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64" names="716">
+				<port id="5" precision="FP32" names="739">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="736" name="Constant_2655" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
+			<output>
+				<port id="0" precision="I64" names="740">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="757" name="__module.model.gpt_neox.layers.2.attention/aten::expand/Abs" type="Abs" version="opset1">
+		<layer id="737" name="__module.model.gpt_neox.layers.3.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
 			<input>
-				<port id="0" precision="I64">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
 			</input>
+			<output>
+				<port id="2" precision="FP32" names="741">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="738" name="__module.model.gpt_neox.layers.3.attention/aten::cat/Concat_5" type="Concat" version="opset1">
+			<data axis="-2" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="present.3.value">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="739" name="Constant_3037" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="740" name="ShapeOf_38925" type="ShapeOf" version="opset3">
+			<data output_type="i64" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+			</input>
 			<output>
 				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="758" name="Maximum_290352" type="Maximum" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="741" name="Constant_38926" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116168" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="742" name="Constant_38927" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64" />
+			</output>
+		</layer>
+		<layer id="743" name="Gather_38928" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="I64">
 					<dim>4</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+				<port id="3" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="759" name="__module.model.gpt_neox.layers.2.attention/aten::add/Add_2" type="Broadcast" version="opset3">
-			<data mode="numpy" />
+		<layer id="744" name="Add_38931" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="I64" names="788">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="745" name="Constant_3040" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="746" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Slice_14" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -11522,19 +11086,28 @@
 					<dim>-1</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="718">
-					<dim>-1</dim>
+				<port id="5" precision="FP32" names="792">
 					<dim>-1</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
 					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="760" name="__module.model.gpt_neox.layers.2.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+		<layer id="747" name="__module.model.gpt_neox.layers.3.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
 			<data causal="false" />
 			<input>
 				<port id="0" precision="FP32">
@@ -11557,13 +11130,13 @@
 				</port>
 				<port id="3" precision="FP32">
 					<dim>-1</dim>
-					<dim>-1</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
 					<dim>-1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="719,720,sdpa_result.5,tensor.9">
+				<port id="4" precision="FP32" names="793,attn_output.19">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -11571,15 +11144,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="761" name="Constant_1854" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="748" name="__module.model.gpt_neox.layers.3.attention/aten::transpose/Constant" type="Const" version="opset1">
+			<data element_type="i32" shape="4" offset="201116200" size="16" />
 			<output>
-				<port id="0" precision="I64" names="721">
+				<port id="0" precision="I32">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="762" name="__module.model.gpt_neox.layers.2.attention/aten::permute/Transpose_3" type="Transpose" version="opset1">
+		<layer id="749" name="__module.model.gpt_neox.layers.3.attention/aten::transpose/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -11587,12 +11160,12 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="I32">
 					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="722">
+				<port id="2" precision="FP32" names="794">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -11600,18 +11173,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="763" name="__module.model.gpt_neox.layers.2.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438650" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="750" name="Constant_40987" type="Const" version="opset1">
+			<data element_type="i64" shape="3" offset="201116216" size="24" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>3</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="764" name="__module.model.gpt_neox.layers.2.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
+		<layer id="751" name="__module.model.gpt_neox.layers.3.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
 			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -11625,15 +11195,15 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="727">
+				<port id="2" precision="FP32" names="797">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="765" name="self.model.gpt_neox.layers.2.attention.dense.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 2560" offset="362996434" size="6553600" />
+		<layer id="752" name="self.model.gpt_neox.layers.3.attention.dense.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 2560" offset="437575760" size="6553600" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>2560</dim>
@@ -11641,7 +11211,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="766" name="Convert_21514" type="Convert" version="opset1">
+		<layer id="753" name="Convert_75209" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -11656,8 +11226,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="767" name="self.model.gpt_neox.layers.2.attention.dense.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="369550034" size="2560" />
+		<layer id="754" name="self.model.gpt_neox.layers.3.attention.dense.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="444129360" size="2560" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>2560</dim>
@@ -11665,7 +11235,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="768" name="Convert_21517" type="Convert" version="opset1">
+		<layer id="755" name="Convert_75212" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -11680,7 +11250,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="769" name="self.model.gpt_neox.layers.2.attention.dense.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="756" name="self.model.gpt_neox.layers.3.attention.dense.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -11699,8 +11269,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="770" name="self.model.gpt_neox.layers.2.attention.dense.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="369552594" size="5120" />
+		<layer id="757" name="self.model.gpt_neox.layers.3.attention.dense.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="444131920" size="5120" />
 			<output>
 				<port id="0" precision="FP16">
 					<dim>2560</dim>
@@ -11708,7 +11278,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="771" name="self.model.gpt_neox.layers.2.attention.dense.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="758" name="self.model.gpt_neox.layers.3.attention.dense.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -11727,11 +11297,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="772" name="self.model.gpt_neox.layers.2.attention.dense.weight" type="Convert" version="opset1">
+		<layer id="759" name="self.model.gpt_neox.layers.3.attention.dense.weight/fq_weights_1/convert" type="Convert" version="opset1">
 			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
 			<input>
 				<port id="0" precision="FP16">
 					<dim>2560</dim>
@@ -11745,7 +11312,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="773" name="__module.model.gpt_neox.layers.2.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="760" name="__module.model.gpt_neox.layers.3.attention.dense/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -11766,37 +11333,17 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="774" name="Constant_299098_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="369557714" size="5120" />
+		<layer id="761" name="Constant_37319" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="444137040" size="10240" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="775" name="Constant_299098" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="776" name="__module.model.gpt_neox.layers.2.attention.dense/aten::linear/Add" type="Add" version="opset1">
+		<layer id="762" name="__module.model.gpt_neox.layers.3.attention.dense/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -11811,14 +11358,14 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="730,732,input.11">
+				<port id="2" precision="FP32" names="800,802,input.13">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="777" name="__module.model.gpt_neox.layers.2/aten::add/Add" type="Add" version="opset1">
+		<layer id="763" name="__module.model.gpt_neox.layers.3/aten::add/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -11833,14 +11380,14 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="750">
+				<port id="2" precision="FP32" names="820">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="778" name="__module.model.gpt_neox.layers.2/aten::add/Add_1" type="Add" version="opset1">
+		<layer id="764" name="__module.model.gpt_neox.layers.3/aten::add/Add_1" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -11855,22 +11402,22 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="751,753,hidden_states.7">
+				<port id="2" precision="FP32" names="821,823,hidden_states.19">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="779" name="__module.model.gpt_neox.layers.3.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867670" size="4" />
+		<layer id="765" name="__module.model.gpt_neox.layers.4.input_layernorm/aten::layer_norm/Multiply" type="Const" version="opset1">
+			<data element_type="i32" shape="1" offset="128867668" size="4" />
 			<output>
 				<port id="0" precision="I32">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="780" name="__module.model.gpt_neox.layers.3.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
+		<layer id="766" name="__module.model.gpt_neox.layers.4.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
 			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
 			<input>
 				<port id="0" precision="FP32">
@@ -11890,37 +11437,17 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="781" name="Constant_299099_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="369562834" size="5120" />
+		<layer id="767" name="Constant_37320" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="444147280" size="10240" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="782" name="Constant_299099" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="783" name="__module.model.gpt_neox.layers.3.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
+		<layer id="768" name="__module.model.gpt_neox.layers.4.post_attention_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -11942,37 +11469,17 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="784" name="Constant_299100_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="369567954" size="5120" />
+		<layer id="769" name="Constant_37321" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="444157520" size="10240" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="785" name="Constant_299100" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="786" name="__module.model.gpt_neox.layers.3.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+		<layer id="770" name="__module.model.gpt_neox.layers.4.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -11987,15 +11494,15 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="888">
+				<port id="2" precision="FP32" names="921">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="787" name="self.model.gpt_neox.layers.3.mlp.dense_h_to_4h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 2560" offset="369573074" size="26214400" />
+		<layer id="771" name="self.model.gpt_neox.layers.4.mlp.dense_h_to_4h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 2560" offset="444167760" size="26214400" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>10240</dim>
@@ -12003,7 +11510,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="788" name="Convert_21564" type="Convert" version="opset1">
+		<layer id="772" name="Convert_74264" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -12018,8 +11525,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="789" name="self.model.gpt_neox.layers.3.mlp.dense_h_to_4h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 1" offset="395787474" size="10240" />
+		<layer id="773" name="self.model.gpt_neox.layers.4.mlp.dense_h_to_4h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 1" offset="470382160" size="10240" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>10240</dim>
@@ -12027,7 +11534,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="790" name="Convert_21567" type="Convert" version="opset1">
+		<layer id="774" name="Convert_74267" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -12042,7 +11549,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="791" name="self.model.gpt_neox.layers.3.mlp.dense_h_to_4h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="775" name="self.model.gpt_neox.layers.4.mlp.dense_h_to_4h.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -12061,8 +11568,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="792" name="self.model.gpt_neox.layers.3.mlp.dense_h_to_4h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="10240, 1" offset="395797714" size="20480" />
+		<layer id="776" name="self.model.gpt_neox.layers.4.mlp.dense_h_to_4h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="10240, 1" offset="470392400" size="20480" />
 			<output>
 				<port id="0" precision="FP16">
 					<dim>10240</dim>
@@ -12070,7 +11577,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="793" name="self.model.gpt_neox.layers.3.mlp.dense_h_to_4h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="777" name="self.model.gpt_neox.layers.4.mlp.dense_h_to_4h.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -12089,7 +11596,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="794" name="self.model.gpt_neox.layers.3.mlp.dense_h_to_4h.weight" type="Convert" version="opset1">
+		<layer id="778" name="__module.model.gpt_neox.layers.4.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
@@ -12107,7 +11614,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="795" name="__module.model.gpt_neox.layers.3.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="779" name="__module.model.gpt_neox.layers.4.mlp.dense_h_to_4h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -12128,37 +11635,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="796" name="Constant_299101_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 10240" offset="395818194" size="20480" />
+		<layer id="780" name="self.model.gpt_neox.layers.4.mlp.dense_h_to_4h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="10240" offset="470412880" size="20480" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.4.mlp.dense_h_to_4h.bias">
 					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="797" name="Constant_299101" type="Convert" version="opset1">
+		<layer id="781" name="__module.model.gpt_neox.layers.4.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
 			</rt_info>
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16">
 					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="798" name="__module.model.gpt_neox.layers.3.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1">
+		<layer id="782" name="__module.model.gpt_neox.layers.4.mlp.dense_h_to_4h/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -12167,20 +11668,18 @@
 					<dim>10240</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="893">
+				<port id="2" precision="FP32" names="926">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="799" name="__module.model.gpt_neox.layers.3.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
+		<layer id="783" name="__module.model.gpt_neox.layers.4.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
 			<data approximation_mode="ERF" />
 			<input>
 				<port id="0" precision="FP32">
@@ -12190,15 +11689,15 @@
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="894">
+				<port id="1" precision="FP32" names="927">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="800" name="self.model.gpt_neox.layers.3.mlp.dense_4h_to_h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 10240" offset="395838674" size="26214400" />
+		<layer id="784" name="self.model.gpt_neox.layers.4.mlp.dense_4h_to_h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 10240" offset="470433360" size="26214400" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>2560</dim>
@@ -12206,7 +11705,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="801" name="Convert_21574" type="Convert" version="opset1">
+		<layer id="785" name="Convert_74274" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -12221,8 +11720,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="802" name="self.model.gpt_neox.layers.3.mlp.dense_4h_to_h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="422053074" size="2560" />
+		<layer id="786" name="self.model.gpt_neox.layers.4.mlp.dense_4h_to_h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="496647760" size="2560" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>2560</dim>
@@ -12230,7 +11729,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="803" name="Convert_21577" type="Convert" version="opset1">
+		<layer id="787" name="Convert_74277" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -12245,7 +11744,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="804" name="self.model.gpt_neox.layers.3.mlp.dense_4h_to_h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="788" name="self.model.gpt_neox.layers.4.mlp.dense_4h_to_h.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -12264,8 +11763,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="805" name="self.model.gpt_neox.layers.3.mlp.dense_4h_to_h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="422055634" size="5120" />
+		<layer id="789" name="self.model.gpt_neox.layers.4.mlp.dense_4h_to_h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="496650320" size="5120" />
 			<output>
 				<port id="0" precision="FP16">
 					<dim>2560</dim>
@@ -12273,7 +11772,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="806" name="self.model.gpt_neox.layers.3.mlp.dense_4h_to_h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="790" name="self.model.gpt_neox.layers.4.mlp.dense_4h_to_h.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -12292,7 +11791,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="807" name="self.model.gpt_neox.layers.3.mlp.dense_4h_to_h.weight" type="Convert" version="opset1">
+		<layer id="791" name="__module.model.gpt_neox.layers.4.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
@@ -12310,7 +11809,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="808" name="__module.model.gpt_neox.layers.3.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="792" name="__module.model.gpt_neox.layers.4.mlp.dense_4h_to_h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -12331,37 +11830,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="809" name="Constant_299102_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="422060754" size="5120" />
+		<layer id="793" name="self.model.gpt_neox.layers.4.mlp.dense_4h_to_h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="2560" offset="496655440" size="5120" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.4.mlp.dense_4h_to_h.bias">
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="810" name="Constant_299102" type="Convert" version="opset1">
+		<layer id="794" name="__module.model.gpt_neox.layers.4.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
 			</rt_info>
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16">
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="811" name="__module.model.gpt_neox.layers.3.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1">
+		<layer id="795" name="__module.model.gpt_neox.layers.4.mlp.dense_4h_to_h/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -12370,50 +11863,28 @@
 					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="897,input.17">
+				<port id="2" precision="FP32" names="930,input.19">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="812" name="Constant_299103_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="422065874" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="813" name="Constant_299103" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="796" name="Constant_37322" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="496660560" size="10240" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="814" name="__module.model.gpt_neox.layers.3.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
+		<layer id="797" name="__module.model.gpt_neox.layers.4.input_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -12435,37 +11906,17 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="815" name="Constant_299104_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="422070994" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="816" name="Constant_299104" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="798" name="Constant_37323" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="496670800" size="10240" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="817" name="__module.model.gpt_neox.layers.3.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+		<layer id="799" name="__module.model.gpt_neox.layers.4.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -12480,15 +11931,15 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="763">
+				<port id="2" precision="FP32" names="833,hidden_states.17">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="818" name="self.model.gpt_neox.layers.3.attention.query_key_value.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 2560" offset="422076114" size="19660800" />
+		<layer id="800" name="self.model.gpt_neox.layers.4.attention.query_key_value.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 2560" offset="496681040" size="19660800" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>7680</dim>
@@ -12496,7 +11947,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="819" name="Convert_21544" type="Convert" version="opset1">
+		<layer id="801" name="Convert_74868" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -12511,8 +11962,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="820" name="self.model.gpt_neox.layers.3.attention.query_key_value.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 1" offset="441736914" size="7680" />
+		<layer id="802" name="self.model.gpt_neox.layers.4.attention.query_key_value.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 1" offset="516341840" size="7680" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>7680</dim>
@@ -12520,7 +11971,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="821" name="Convert_21547" type="Convert" version="opset1">
+		<layer id="803" name="Convert_74871" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -12535,7 +11986,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="822" name="self.model.gpt_neox.layers.3.attention.query_key_value.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="804" name="self.model.gpt_neox.layers.4.attention.query_key_value.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -12554,8 +12005,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="823" name="self.model.gpt_neox.layers.3.attention.query_key_value.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="7680, 1" offset="441744594" size="15360" />
+		<layer id="805" name="self.model.gpt_neox.layers.4.attention.query_key_value.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="7680, 1" offset="516349520" size="15360" />
 			<output>
 				<port id="0" precision="FP16">
 					<dim>7680</dim>
@@ -12563,7 +12014,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="824" name="self.model.gpt_neox.layers.3.attention.query_key_value.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="806" name="self.model.gpt_neox.layers.4.attention.query_key_value.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -12582,11 +12033,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="825" name="self.model.gpt_neox.layers.3.attention.query_key_value.weight" type="Convert" version="opset1">
+		<layer id="807" name="self.model.gpt_neox.layers.4.attention.query_key_value.weight/fq_weights_1/convert" type="Convert" version="opset1">
 			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
 			<input>
 				<port id="0" precision="FP16">
 					<dim>7680</dim>
@@ -12600,7 +12048,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="826" name="__module.model.gpt_neox.layers.3.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="808" name="__module.model.gpt_neox.layers.4.attention.query_key_value/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -12621,37 +12069,17 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="827" name="Constant_299105_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 7680" offset="441759954" size="15360" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="828" name="Constant_299105" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</input>
+		<layer id="809" name="Constant_37324" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 7680" offset="516364880" size="30720" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="829" name="__module.model.gpt_neox.layers.3.attention.query_key_value/aten::linear/Add" type="Add" version="opset1">
+		<layer id="810" name="__module.model.gpt_neox.layers.4.attention.query_key_value/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -12666,25 +12094,22 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="770,qkv.13">
+				<port id="2" precision="FP32" names="840,qkv.17">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="830" name="__module.model.gpt_neox.layers.3.attention/prim::ListConstruct/Concat" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080154" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="811" name="__module.model.gpt_neox.layers.4.attention/prim::ListConstruct" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201115992" size="32" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="831" name="__module.model.gpt_neox.layers.3.attention/aten::view/Reshape" type="Reshape" version="opset1">
+		<layer id="812" name="__module.model.gpt_neox.layers.4.attention/aten::view/Reshape" type="Reshape" version="opset1">
 			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -12697,7 +12122,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="774,qkv.15">
+				<port id="2" precision="FP32" names="844,qkv.19">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -12705,41 +12130,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="832" name="Constant_270154" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="813" name="Constant_3236" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="833" name="Constant_270157" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="814" name="Constant_3238" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="834" name="Constant_270160" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="815" name="Constant_3240" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="835" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="816" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Slice" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -12748,17 +12163,20 @@
 					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="775">
+				<port id="5" precision="FP32" names="845">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -12766,15 +12184,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="836" name="Constant_1966" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="817" name="Constant_3249" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64" names="776">
+				<port id="0" precision="I64" names="846">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="837" name="__module.model.gpt_neox.layers.3.attention/aten::permute/Transpose" type="Transpose" version="opset1">
+		<layer id="818" name="__module.model.gpt_neox.layers.4.attention/aten::permute/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -12787,7 +12205,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="777,query.19">
+				<port id="2" precision="FP32" names="847,query.9">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -12795,41 +12213,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="838" name="Constant_270166" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="819" name="Constant_3290" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="839" name="Constant_270169" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="820" name="Constant_3292" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="840" name="Constant_270172" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="821" name="Constant_3294" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="841" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="822" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Slice_1" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -12838,17 +12246,20 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="784,q.7">
+				<port id="5" precision="FP32" names="854,q.9">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -12856,215 +12267,360 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="842" name="self.model.gpt_neox.layers.3.attention.rotary_emb.cos_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201080346" size="81920" />
+		<layer id="823" name="__module.model.gpt_neox.layers.4.attention/aten::mul/Multiply" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.3.attention.rotary_emb.cos_cached">
-					<dim>2048</dim>
+				<port id="2" precision="FP32" names="860">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="843" name="self.model.gpt_neox.layers.3.attention.rotary_emb.cos_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="824" name="Constant_3448" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="825" name="Constant_3450" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="826" name="Constant_3452" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="827" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Slice_2" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="5" precision="FP32" names="870,x2.17">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="844" name="__module.model.gpt_neox.layers.3.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="828" name="Constant_37325" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="845" name="Constant_270178" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="829" name="__module.model.gpt_neox.layers.4.attention/aten::neg/Multiply" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="871">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="830" name="Constant_3345" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="846" name="Constant_270181" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="831" name="Constant_3389" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="847" name="Constant_270184" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="832" name="Constant_3391" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="848" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="833" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Slice_3" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>240</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="778">
-					<dim>-1</dim>
+				<port id="5" precision="FP32" names="865">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="849" name="Constant_1980" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="834" name="__module.model.gpt_neox.layers.4.attention/aten::cat/Concat" type="Concat" version="opset1">
+			<data axis="-1" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64" names="779">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="873">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="850" name="__module.model.gpt_neox.layers.3.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
+		<layer id="835" name="__module.model.gpt_neox.layers.4.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="780,key.19">
+				<port id="2" precision="FP32" names="874">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="851" name="__module.model.gpt_neox.layers.3.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="836" name="__module.model.gpt_neox.layers.4.attention/aten::add/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="875">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="852" name="__module.model.gpt_neox.layers.3.attention/aten::size/Constant_2" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="837" name="Constant_3300" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="853" name="__module.model.gpt_neox.layers.3.attention/aten::size/Gather_2" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
+		<layer id="838" name="Constant_3302" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
+			</output>
+		</layer>
+		<layer id="839" name="Constant_3304" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="3" precision="I64" names="788" />
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="854" name="Constant_337507_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
+		<layer id="840" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Slice_4" type="Slice" version="opset8">
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="FP16" />
+				<port id="5" precision="FP32" names="855">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="855" name="Constant_337507" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="841" name="__module.model.gpt_neox.layers.4.attention/aten::cat/Concat_1" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" />
+				<port id="2" precision="FP32" names="893">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="842" name="Constant_52647" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
+			<output>
+				<port id="0" precision="FP32" />
 			</output>
 		</layer>
-		<layer id="856" name="Constant_337500" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
+		<layer id="843" name="Constant_52640" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="857" name="Constant_337502" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
+		<layer id="844" name="Constant_52642" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="858" name="Constant_337504" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
+		<layer id="845" name="Constant_52644" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="859" name="Concat_337505" type="Concat" version="opset1">
+		<layer id="846" name="Concat_52645" type="Concat" version="opset1">
 			<data axis="0" />
 			<input>
 				<port id="0" precision="I64">
@@ -13086,7 +12642,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="860" name="Broadcast_337508" type="Broadcast" version="opset3">
+		<layer id="847" name="Broadcast_52648" type="Broadcast" version="opset3">
 			<data mode="numpy" />
 			<input>
 				<port id="0" precision="FP32" />
@@ -13103,8 +12659,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="861" name="ReadValue_330517" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.3.keypresent.3.key" variable_type="f32" variable_shape="?,32,?,80" />
+		<layer id="848" name="ReadValue_50287" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.4.keypresent.4.key" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -13114,7 +12670,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="past_key_values.3.key">
+				<port id="1" precision="FP32" names="past_key_values.4.key">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -13122,13 +12678,13 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="862" name="Constant_323403" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
+		<layer id="849" name="Constant_47802" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="863" name="Gather_323404" type="Gather" version="opset8">
+		<layer id="850" name="Gather_47803" type="Gather" version="opset8">
 			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="FP32">
@@ -13151,91 +12707,37 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="864" name="__module.model.gpt_neox.layers.3.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="865" name="__module.model.gpt_neox.layers.3.attention/aten::size/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="851" name="Constant_3254" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="866" name="__module.model.gpt_neox.layers.3.attention/aten::size/Gather_3" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="790" />
-			</output>
-		</layer>
-		<layer id="867" name="__module.model.gpt_neox.layers.3.attention/aten::add_/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" names="789,793,794,seq_len.13" />
-			</output>
-		</layer>
-		<layer id="868" name="__module.model.gpt_neox.layers.3.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
 			</output>
 		</layer>
-		<layer id="869" name="__module.model.gpt_neox.layers.3.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="852" name="Constant_3256" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="870" name="__module.model.gpt_neox.layers.3.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="853" name="Constant_3258" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="871" name="__module.model.gpt_neox.layers.3.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="854" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Slice_5" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -13246,60 +12748,103 @@
 				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="797,800,cos.13">
+				<port id="5" precision="FP32" names="848">
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="872" name="Constant_38215" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
+		<layer id="855" name="Constant_3267" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="I64" names="849">
+					<dim>4</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="873" name="__module.model.gpt_neox.layers.3.attention/aten::index/Gather" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="856" name="__module.model.gpt_neox.layers.4.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="803">
+				<port id="2" precision="FP32" names="850,key.17">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="857" name="Constant_3310" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="858" name="Constant_3312" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="859" name="Constant_3314" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="874" name="__module.model.gpt_neox.layers.3.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1">
+		<layer id="860" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Slice_6" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="804,cos.15">
+				<port id="5" precision="FP32" names="856,k.9">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="875" name="__module.model.gpt_neox.layers.3.attention/aten::mul/Multiply" type="Multiply" version="opset1">
+		<layer id="861" name="__module.model.gpt_neox.layers.4.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -13316,7 +12861,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="808">
+				<port id="2" precision="FP32" names="876">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -13324,30 +12869,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="876" name="Constant_270189" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="862" name="Constant_3591" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="877" name="Constant_270188" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="863" name="Constant_3593" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="878" name="__module.model.gpt_neox.layers.3.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="864" name="Constant_3595" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="865" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Slice_7" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -13355,148 +12901,21 @@
 					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
-			</input>
-			<output>
 				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="879" name="__module.model.gpt_neox.layers.3.attention/aten::size/Constant_5" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="880" name="__module.model.gpt_neox.layers.3.attention/aten::size/Gather_5" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="809,814" />
-			</output>
-		</layer>
-		<layer id="881" name="__module.model.gpt_neox.layers.3.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" />
-			</output>
-		</layer>
-		<layer id="882" name="__module.model.gpt_neox.layers.3.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-			</input>
-			<output>
-				<port id="1" precision="I64" names="811,812,816,817" />
-			</output>
-		</layer>
-		<layer id="883" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Constant_1" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="884" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="885" name="Constant_270187" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="886" name="ScatterUpdate_270190" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
 				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="887" name="Constant_270193" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="888" name="Constant_270196" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="889" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="818,x2.13">
+				<port id="5" precision="FP32" names="886,x2.19">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -13504,32 +12923,10 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="890" name="Constant_299106_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="891" name="Constant_299106" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
+		<layer id="866" name="Constant_37326" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>1</dim>
@@ -13537,7 +12934,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="892" name="__module.model.gpt_neox.layers.3.attention/aten::neg/Multiply" type="Multiply" version="opset1">
+		<layer id="867" name="__module.model.gpt_neox.layers.4.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -13554,7 +12951,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="819">
+				<port id="2" precision="FP32" names="887">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -13562,30 +12959,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="893" name="Constant_270241" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="868" name="Constant_3488" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="894" name="Constant_270246" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="869" name="Constant_3532" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="870" name="Constant_3534" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="895" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="871" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Slice_8" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -13594,17 +12992,20 @@
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="813">
+				<port id="5" precision="FP32" names="881">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -13612,7 +13013,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="896" name="__module.model.gpt_neox.layers.3.attention/aten::cat/Concat" type="Concat" version="opset1">
+		<layer id="872" name="__module.model.gpt_neox.layers.4.attention/aten::cat/Concat_2" type="Concat" version="opset1">
 			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
@@ -13629,7 +13030,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="821">
+				<port id="2" precision="FP32" names="889">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -13637,61 +13038,87 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="897" name="self.model.gpt_neox.layers.3.attention.rotary_emb.sin_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201162356" size="81920" />
+		<layer id="873" name="__module.model.gpt_neox.layers.4.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.3.attention.rotary_emb.sin_cached">
-					<dim>2048</dim>
+				<port id="2" precision="FP32" names="890">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="898" name="self.model.gpt_neox.layers.3.attention.rotary_emb.sin_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="874" name="__module.model.gpt_neox.layers.4.attention/aten::add/Add_1" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
+				<port id="2" precision="FP32" names="891">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="899" name="__module.model.gpt_neox.layers.3.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="875" name="Constant_3320" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="900" name="__module.model.gpt_neox.layers.3.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="876" name="Constant_3322" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="877" name="Constant_3324" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="901" name="__module.model.gpt_neox.layers.3.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="878" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Slice_9" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -13702,61 +13129,21 @@
 				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="798,801,sin.13">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="902" name="Constant_38218" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="903" name="__module.model.gpt_neox.layers.3.attention/aten::index/Gather_1" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="806">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="904" name="__module.model.gpt_neox.layers.3.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="807,sin.15">
+				<port id="5" precision="FP32" names="857">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>60</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="905" name="__module.model.gpt_neox.layers.3.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="879" name="__module.model.gpt_neox.layers.4.attention/aten::cat/Concat_3" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -13766,124 +13153,155 @@
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>60</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="822">
+				<port id="2" precision="FP32" names="895">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="906" name="__module.model.gpt_neox.layers.3.attention/aten::add/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="880" name="__module.model.gpt_neox.layers.4.attention/aten::cat/Concat_4" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="823">
+				<port id="2" precision="FP32" names="present.4.key">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="907" name="Constant_270292" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="881" name="Constant_52659" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
+			<output>
+				<port id="0" precision="FP32" />
+			</output>
+		</layer>
+		<layer id="882" name="Constant_52652" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="908" name="Constant_270295" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="883" name="Constant_52654" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="909" name="Constant_270298" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="884" name="Constant_52656" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="910" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="885" name="Concat_52657" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="4" precision="I64">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="886" name="Broadcast_52660" type="Broadcast" version="opset3">
+			<data mode="numpy" />
+			<input>
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="785">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="911" name="__module.model.gpt_neox.layers.3.attention/aten::cat/Concat_1" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="887" name="ReadValue_50289" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.4.valuepresent.4.value" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
 				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP32" names="past_key_values.4.value">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
+			</output>
+		</layer>
+		<layer id="888" name="Constant_47805" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64" />
+			</output>
+		</layer>
+		<layer id="889" name="Gather_47806" type="Gather" version="opset8">
+			<data batch_dims="0" />
+			<input>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>60</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I32">
+					<dim>-1</dim>
 				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="841,851,query.21,query.23">
+				<port id="3" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -13891,122 +13309,130 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="912" name="Constant_270304" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="890" name="Constant_3272" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="913" name="Constant_270307" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="891" name="Constant_3274" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="914" name="Constant_270310" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="892" name="Constant_3276" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="915" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="893" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Slice_10" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="5" precision="FP32" names="851">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="894" name="Constant_3285" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
+			<output>
+				<port id="0" precision="I64" names="852">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="895" name="__module.model.gpt_neox.layers.4.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="786,k.7">
+				<port id="2" precision="FP32" names="853">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="916" name="__module.model.gpt_neox.layers.3.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="896" name="__module.model.gpt_neox.layers.4.attention/aten::cat/Concat_5" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="824">
+				<port id="2" precision="FP32" names="present.4.value">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="917" name="Constant_270314" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="918" name="Constant_270313" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="897" name="Constant_3667" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="919" name="__module.model.gpt_neox.layers.3.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3">
+		<layer id="898" name="ShapeOf_38996" type="ShapeOf" version="opset3">
 			<data output_type="i64" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
@@ -14015,81 +13441,68 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="920" name="__module.model.gpt_neox.layers.3.attention/aten::size/Constant_7" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="899" name="Constant_38997" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116168" size="8" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="921" name="__module.model.gpt_neox.layers.3.attention/aten::size/Gather_7" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="825,830" />
 			</output>
 		</layer>
-		<layer id="922" name="__module.model.gpt_neox.layers.3.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
+		<layer id="900" name="Constant_38998" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="2" precision="I64" />
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="923" name="__module.model.gpt_neox.layers.3.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1">
+		<layer id="901" name="Gather_38999" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
-				<port id="0" precision="I64" />
+				<port id="0" precision="I64">
+					<dim>4</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="1" precision="I64" names="827,828,832,833" />
-			</output>
-		</layer>
-		<layer id="924" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="925" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1">
+		<layer id="902" name="Add_39002" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
+				<port id="2" precision="I64" names="900">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="926" name="Constant_270312" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="903" name="Constant_3670" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="927" name="ScatterUpdate_270315" type="ScatterUpdate" version="opset3">
+		<layer id="904" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Slice_14" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -14097,37673 +13510,81 @@
 				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
 				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="928" name="Constant_270318" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="929" name="Constant_270321" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="5" precision="FP32" names="904">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="930" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="905" name="__module.model.gpt_neox.layers.4.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+			<data causal="false" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="3" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="834,x2.15">
+				<port id="4" precision="FP32" names="905,attn_output.25">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="931" name="Constant_299107_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
+		<layer id="906" name="__module.model.gpt_neox.layers.4.attention/aten::transpose/Constant" type="Const" version="opset1">
+			<data element_type="i32" shape="4" offset="201116200" size="16" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="932" name="Constant_299107" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="933" name="__module.model.gpt_neox.layers.3.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="835">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="934" name="Constant_270366" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="935" name="Constant_270371" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="936" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="829">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="937" name="__module.model.gpt_neox.layers.3.attention/aten::cat/Concat_2" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="837">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="938" name="__module.model.gpt_neox.layers.3.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="838">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="939" name="__module.model.gpt_neox.layers.3.attention/aten::add/Add_1" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="839">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="940" name="Constant_270416" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="941" name="Constant_270419" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="942" name="Constant_270422" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="943" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="787">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="944" name="__module.model.gpt_neox.layers.3.attention/aten::cat/Concat_3" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="843">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="945" name="__module.model.gpt_neox.layers.3.attention/aten::to/Convert_1" type="Concat" version="opset1">
-			<data axis="-2" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="present.3.key">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="946" name="Constant_337520_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="947" name="Constant_337520" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="948" name="Constant_337513" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="949" name="Constant_337515" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="950" name="Constant_337517" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="951" name="Concat_337518" type="Concat" version="opset1">
-			<data axis="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="952" name="Broadcast_337521" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="953" name="ReadValue_330519" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.3.valuepresent.3.value" variable_type="f32" variable_shape="?,32,?,80" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32" names="past_key_values.3.value">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="954" name="Constant_323406" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64" />
-			</output>
-		</layer>
-		<layer id="955" name="Gather_323407" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I64" />
-			</input>
-			<output>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="956" name="Constant_270428" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="957" name="Constant_270431" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="958" name="Constant_270434" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="959" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="781">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="960" name="Constant_1994" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="782">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="961" name="__module.model.gpt_neox.layers.3.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="783">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="962" name="__module.model.gpt_neox.layers.3.attention/aten::cat/Concat_5" type="Concat" version="opset1">
-			<data axis="-2" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="present.3.value">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="963" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Slice_11" type="Const" version="opset1">
-			<data element_type="boolean" shape="1, 1, 2048, 2048" offset="201244276" size="4194304" />
-			<output>
-				<port id="0" precision="BOOL" names="860">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="964" name="Constant_270438" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="965" name="Constant_270437" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162298" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="966" name="__module.model.gpt_neox.layers.3.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="967" name="__module.model.gpt_neox.layers.3.attention/aten::size/Constant_8" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="968" name="__module.model.gpt_neox.layers.3.attention/aten::size/Gather_8" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="855" />
-			</output>
-		</layer>
-		<layer id="969" name="__module.model.gpt_neox.layers.3.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="970" name="__module.model.gpt_neox.layers.3.attention/aten::size/Constant_9" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="971" name="__module.model.gpt_neox.layers.3.attention/aten::size/Gather_9" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="853" />
-			</output>
-		</layer>
-		<layer id="972" name="__module.model.gpt_neox.layers.3.attention/aten::sub/Subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" names="857,858" />
-			</output>
-		</layer>
-		<layer id="973" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Constant_4" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="974" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="975" name="Constant_270436" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="976" name="ScatterUpdate_270439" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="977" name="Constant_270440" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="978" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="979" name="ScatterUpdate_270441" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="980" name="Constant_270444" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438612" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="981" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>3</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="BOOL" names="861">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="982" name="Constant_270517" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="983" name="Constant_270518" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="984" name="Constant_270514" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="985" name="Constant_270513" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="986" name="ScatterUpdate_270519" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="987" name="Constant_270522" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="988" name="__module.model.gpt_neox.layers.3.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="BOOL" names="862,863">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="989" name="__module.model.gpt_neox.layers.3.attention/aten::where/Convert_1_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="990" name="__module.model.gpt_neox.layers.3.attention/aten::where/Convert_1" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="991" name="__module.model.gpt_neox.layers.3.attention/aten::full/Convert" type="Const" version="opset1">
-			<data element_type="f32" shape="" offset="205438636" size="4" />
-			<output>
-				<port id="0" precision="FP32" names="850" />
-			</output>
-		</layer>
-		<layer id="992" name="__module.model.gpt_neox.layers.3.attention/aten::where/Select" type="Select" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="1" precision="FP32" />
-				<port id="2" precision="FP32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="864,causal_mask.13">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="993" name="Add_290356" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="994" name="ShapeOf_290357" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="995" name="Constant_254374" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="996" name="__module.model.gpt_neox.layers.3.attention/aten::size/Constant_10" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="997" name="__module.model.gpt_neox.layers.3.attention/aten::size/Gather_10" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="848">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="998" name="__module.model.gpt_neox.layers.3.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1">
-			<data axis="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64" names="865">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="999" name="__module.model.gpt_neox.layers.3.attention/aten::expand/Abs" type="Abs" version="opset1">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1000" name="Maximum_290358" type="Maximum" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1001" name="__module.model.gpt_neox.layers.3.attention/aten::add/Add_2" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="867">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1002" name="__module.model.gpt_neox.layers.3.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
-			<data causal="false" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="868,869,sdpa_result.7,tensor.13">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1003" name="Constant_2435" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="870">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1004" name="__module.model.gpt_neox.layers.3.attention/aten::permute/Transpose_3" type="Transpose" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="871">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1005" name="__module.model.gpt_neox.layers.3.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438650" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1006" name="__module.model.gpt_neox.layers.3.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
-			<data special_zero="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>3</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="876">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1007" name="self.model.gpt_neox.layers.3.attention.dense.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 2560" offset="441775314" size="6553600" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1008" name="Convert_21554" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1009" name="self.model.gpt_neox.layers.3.attention.dense.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="448328914" size="2560" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1010" name="Convert_21557" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1011" name="self.model.gpt_neox.layers.3.attention.dense.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1012" name="self.model.gpt_neox.layers.3.attention.dense.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="448331474" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1013" name="self.model.gpt_neox.layers.3.attention.dense.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1014" name="self.model.gpt_neox.layers.3.attention.dense.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1015" name="__module.model.gpt_neox.layers.3.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1016" name="Constant_299108_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="448336594" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1017" name="Constant_299108" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1018" name="__module.model.gpt_neox.layers.3.attention.dense/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="879,881,input.15">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1019" name="__module.model.gpt_neox.layers.3/aten::add/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="899">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1020" name="__module.model.gpt_neox.layers.3/aten::add/Add_1" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="900,902,hidden_states.9">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1021" name="__module.model.gpt_neox.layers.4.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867670" size="4" />
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1022" name="__module.model.gpt_neox.layers.4.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
-			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1023" name="Constant_299109_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="448341714" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1024" name="Constant_299109" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1025" name="__module.model.gpt_neox.layers.4.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1026" name="Constant_299110_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="448346834" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1027" name="Constant_299110" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1028" name="__module.model.gpt_neox.layers.4.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1037">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1029" name="self.model.gpt_neox.layers.4.mlp.dense_h_to_4h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 2560" offset="448351954" size="26214400" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1030" name="Convert_21604" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1031" name="self.model.gpt_neox.layers.4.mlp.dense_h_to_4h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 1" offset="474566354" size="10240" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1032" name="Convert_21607" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1033" name="self.model.gpt_neox.layers.4.mlp.dense_h_to_4h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1034" name="self.model.gpt_neox.layers.4.mlp.dense_h_to_4h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="10240, 1" offset="474576594" size="20480" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1035" name="self.model.gpt_neox.layers.4.mlp.dense_h_to_4h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1036" name="self.model.gpt_neox.layers.4.mlp.dense_h_to_4h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1037" name="__module.model.gpt_neox.layers.4.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1038" name="Constant_299111_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 10240" offset="474597074" size="20480" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1039" name="Constant_299111" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1040" name="__module.model.gpt_neox.layers.4.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1042">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1041" name="__module.model.gpt_neox.layers.4.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
-			<data approximation_mode="ERF" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32" names="1043">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1042" name="self.model.gpt_neox.layers.4.mlp.dense_4h_to_h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 10240" offset="474617554" size="26214400" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1043" name="Convert_21614" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1044" name="self.model.gpt_neox.layers.4.mlp.dense_4h_to_h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="500831954" size="2560" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1045" name="Convert_21617" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1046" name="self.model.gpt_neox.layers.4.mlp.dense_4h_to_h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1047" name="self.model.gpt_neox.layers.4.mlp.dense_4h_to_h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="500834514" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1048" name="self.model.gpt_neox.layers.4.mlp.dense_4h_to_h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1049" name="self.model.gpt_neox.layers.4.mlp.dense_4h_to_h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1050" name="__module.model.gpt_neox.layers.4.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1051" name="Constant_299112_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="500839634" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1052" name="Constant_299112" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1053" name="__module.model.gpt_neox.layers.4.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1046,input.21">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1054" name="Constant_299113_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="500844754" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1055" name="Constant_299113" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1056" name="__module.model.gpt_neox.layers.4.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1057" name="Constant_299114_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="500849874" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1058" name="Constant_299114" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1059" name="__module.model.gpt_neox.layers.4.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="912">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1060" name="self.model.gpt_neox.layers.4.attention.query_key_value.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 2560" offset="500854994" size="19660800" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1061" name="Convert_21584" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1062" name="self.model.gpt_neox.layers.4.attention.query_key_value.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 1" offset="520515794" size="7680" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1063" name="Convert_21587" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1064" name="self.model.gpt_neox.layers.4.attention.query_key_value.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1065" name="self.model.gpt_neox.layers.4.attention.query_key_value.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="7680, 1" offset="520523474" size="15360" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1066" name="self.model.gpt_neox.layers.4.attention.query_key_value.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1067" name="self.model.gpt_neox.layers.4.attention.query_key_value.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1068" name="__module.model.gpt_neox.layers.4.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1069" name="Constant_299115_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 7680" offset="520538834" size="15360" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1070" name="Constant_299115" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1071" name="__module.model.gpt_neox.layers.4.attention.query_key_value/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="919,qkv.17">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1072" name="__module.model.gpt_neox.layers.4.attention/prim::ListConstruct/Concat" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080154" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1073" name="__module.model.gpt_neox.layers.4.attention/aten::view/Reshape" type="Reshape" version="opset1">
-			<data special_zero="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="923,qkv.19">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1074" name="Constant_270529" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1075" name="Constant_270532" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1076" name="Constant_270535" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1077" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="924">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1078" name="Constant_2547" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="925">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1079" name="__module.model.gpt_neox.layers.4.attention/aten::permute/Transpose" type="Transpose" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="926,query.25">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1080" name="Constant_270541" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1081" name="Constant_270544" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1082" name="Constant_270547" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1083" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="933,q.9">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1084" name="self.model.gpt_neox.layers.4.attention.rotary_emb.cos_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201080346" size="81920" />
-			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.4.attention.rotary_emb.cos_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1085" name="self.model.gpt_neox.layers.4.attention.rotary_emb.cos_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1086" name="__module.model.gpt_neox.layers.4.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1087" name="Constant_270553" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1088" name="Constant_270556" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1089" name="Constant_270559" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1090" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="927">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1091" name="Constant_2561" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="928">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1092" name="__module.model.gpt_neox.layers.4.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="929,key.25">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1093" name="__module.model.gpt_neox.layers.4.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1094" name="__module.model.gpt_neox.layers.4.attention/aten::size/Constant_2" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1095" name="__module.model.gpt_neox.layers.4.attention/aten::size/Gather_2" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="937" />
-			</output>
-		</layer>
-		<layer id="1096" name="Constant_337533_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="1097" name="Constant_337533" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="1098" name="Constant_337526" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1099" name="Constant_337528" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1100" name="Constant_337530" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1101" name="Concat_337531" type="Concat" version="opset1">
-			<data axis="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1102" name="Broadcast_337534" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1103" name="ReadValue_330529" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.4.keypresent.4.key" variable_type="f32" variable_shape="?,32,?,80" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32" names="past_key_values.4.key">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1104" name="Constant_323409" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64" />
-			</output>
-		</layer>
-		<layer id="1105" name="Gather_323410" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I64" />
-			</input>
-			<output>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1106" name="__module.model.gpt_neox.layers.4.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1107" name="__module.model.gpt_neox.layers.4.attention/aten::size/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1108" name="__module.model.gpt_neox.layers.4.attention/aten::size/Gather_3" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="939" />
-			</output>
-		</layer>
-		<layer id="1109" name="__module.model.gpt_neox.layers.4.attention/aten::add_/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" names="938,942,943,seq_len.17" />
-			</output>
-		</layer>
-		<layer id="1110" name="__module.model.gpt_neox.layers.4.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1111" name="__module.model.gpt_neox.layers.4.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1112" name="__module.model.gpt_neox.layers.4.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1113" name="__module.model.gpt_neox.layers.4.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="946,949,cos.17">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1114" name="Constant_38221" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1115" name="__module.model.gpt_neox.layers.4.attention/aten::index/Gather" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="952">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1116" name="__module.model.gpt_neox.layers.4.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="953,cos.19">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1117" name="__module.model.gpt_neox.layers.4.attention/aten::mul/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="957">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1118" name="Constant_270564" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1119" name="Constant_270563" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1120" name="__module.model.gpt_neox.layers.4.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1121" name="__module.model.gpt_neox.layers.4.attention/aten::size/Constant_5" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1122" name="__module.model.gpt_neox.layers.4.attention/aten::size/Gather_5" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="958,963" />
-			</output>
-		</layer>
-		<layer id="1123" name="__module.model.gpt_neox.layers.4.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" />
-			</output>
-		</layer>
-		<layer id="1124" name="__module.model.gpt_neox.layers.4.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-			</input>
-			<output>
-				<port id="1" precision="I64" names="960,961,965,966" />
-			</output>
-		</layer>
-		<layer id="1125" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Constant_1" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1126" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1127" name="Constant_270562" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1128" name="ScatterUpdate_270565" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1129" name="Constant_270568" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1130" name="Constant_270571" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1131" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="967,x2.17">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1132" name="Constant_299116_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1133" name="Constant_299116" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1134" name="__module.model.gpt_neox.layers.4.attention/aten::neg/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="968">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1135" name="Constant_270616" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1136" name="Constant_270621" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1137" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="962">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1138" name="__module.model.gpt_neox.layers.4.attention/aten::cat/Concat" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="970">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1139" name="self.model.gpt_neox.layers.4.attention.rotary_emb.sin_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201162356" size="81920" />
-			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.4.attention.rotary_emb.sin_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1140" name="self.model.gpt_neox.layers.4.attention.rotary_emb.sin_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1141" name="__module.model.gpt_neox.layers.4.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1142" name="__module.model.gpt_neox.layers.4.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1143" name="__module.model.gpt_neox.layers.4.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="947,950,sin.17">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1144" name="Constant_38224" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1145" name="__module.model.gpt_neox.layers.4.attention/aten::index/Gather_1" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="955">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1146" name="__module.model.gpt_neox.layers.4.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="956,sin.19">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1147" name="__module.model.gpt_neox.layers.4.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="971">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1148" name="__module.model.gpt_neox.layers.4.attention/aten::add/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="972">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1149" name="Constant_270667" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1150" name="Constant_270670" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1151" name="Constant_270673" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1152" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="934">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1153" name="__module.model.gpt_neox.layers.4.attention/aten::cat/Concat_1" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1000,990,query.27,query.29">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1154" name="Constant_270679" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1155" name="Constant_270682" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1156" name="Constant_270685" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1157" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="935,k.9">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1158" name="__module.model.gpt_neox.layers.4.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="973">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1159" name="Constant_270689" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1160" name="Constant_270688" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1161" name="__module.model.gpt_neox.layers.4.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1162" name="__module.model.gpt_neox.layers.4.attention/aten::size/Constant_7" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1163" name="__module.model.gpt_neox.layers.4.attention/aten::size/Gather_7" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="974,979" />
-			</output>
-		</layer>
-		<layer id="1164" name="__module.model.gpt_neox.layers.4.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" />
-			</output>
-		</layer>
-		<layer id="1165" name="__module.model.gpt_neox.layers.4.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-			</input>
-			<output>
-				<port id="1" precision="I64" names="976,977,981,982" />
-			</output>
-		</layer>
-		<layer id="1166" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1167" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1168" name="Constant_270687" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1169" name="ScatterUpdate_270690" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1170" name="Constant_270693" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1171" name="Constant_270696" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1172" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="983,x2.19">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1173" name="Constant_299117_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1174" name="Constant_299117" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1175" name="__module.model.gpt_neox.layers.4.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="984">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1176" name="Constant_270741" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1177" name="Constant_270746" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1178" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="978">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1179" name="__module.model.gpt_neox.layers.4.attention/aten::cat/Concat_2" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="986">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1180" name="__module.model.gpt_neox.layers.4.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="987">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1181" name="__module.model.gpt_neox.layers.4.attention/aten::add/Add_1" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="988">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1182" name="Constant_270791" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1183" name="Constant_270794" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1184" name="Constant_270797" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1185" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="936">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1186" name="__module.model.gpt_neox.layers.4.attention/aten::cat/Concat_3" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="992">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1187" name="__module.model.gpt_neox.layers.4.attention/aten::to/Convert_1" type="Concat" version="opset1">
-			<data axis="-2" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="present.4.key">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1188" name="Constant_337546_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="1189" name="Constant_337546" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="1190" name="Constant_337539" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1191" name="Constant_337541" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1192" name="Constant_337543" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1193" name="Concat_337544" type="Concat" version="opset1">
-			<data axis="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1194" name="Broadcast_337547" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1195" name="ReadValue_330531" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.4.valuepresent.4.value" variable_type="f32" variable_shape="?,32,?,80" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32" names="past_key_values.4.value">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1196" name="Constant_323412" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64" />
-			</output>
-		</layer>
-		<layer id="1197" name="Gather_323413" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I64" />
-			</input>
-			<output>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1198" name="Constant_270803" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1199" name="Constant_270806" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1200" name="Constant_270809" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1201" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="930">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1202" name="Constant_2575" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="931">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1203" name="__module.model.gpt_neox.layers.4.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="932">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1204" name="__module.model.gpt_neox.layers.4.attention/aten::cat/Concat_5" type="Concat" version="opset1">
-			<data axis="-2" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="present.4.value">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1205" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Slice_11" type="Const" version="opset1">
-			<data element_type="boolean" shape="1, 1, 2048, 2048" offset="201244276" size="4194304" />
-			<output>
-				<port id="0" precision="BOOL" names="1009">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1206" name="Constant_270813" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1207" name="Constant_270812" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162298" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1208" name="__module.model.gpt_neox.layers.4.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1209" name="__module.model.gpt_neox.layers.4.attention/aten::size/Constant_8" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1210" name="__module.model.gpt_neox.layers.4.attention/aten::size/Gather_8" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1004" />
-			</output>
-		</layer>
-		<layer id="1211" name="__module.model.gpt_neox.layers.4.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1212" name="__module.model.gpt_neox.layers.4.attention/aten::size/Constant_9" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1213" name="__module.model.gpt_neox.layers.4.attention/aten::size/Gather_9" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1002" />
-			</output>
-		</layer>
-		<layer id="1214" name="__module.model.gpt_neox.layers.4.attention/aten::sub/Subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" names="1006,1007" />
-			</output>
-		</layer>
-		<layer id="1215" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Constant_4" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1216" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1217" name="Constant_270811" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1218" name="ScatterUpdate_270814" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1219" name="Constant_270815" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1220" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1221" name="ScatterUpdate_270816" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1222" name="Constant_270819" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438612" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1223" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>3</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="BOOL" names="1010">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1224" name="Constant_270892" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1225" name="Constant_270893" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1226" name="Constant_270889" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1227" name="Constant_270888" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1228" name="ScatterUpdate_270894" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1229" name="Constant_270897" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1230" name="__module.model.gpt_neox.layers.4.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="BOOL" names="1011,1012">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1231" name="__module.model.gpt_neox.layers.4.attention/aten::where/Convert_1_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="1232" name="__module.model.gpt_neox.layers.4.attention/aten::where/Convert_1" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="1233" name="__module.model.gpt_neox.layers.4.attention/aten::full/Convert" type="Const" version="opset1">
-			<data element_type="f32" shape="" offset="205438636" size="4" />
-			<output>
-				<port id="0" precision="FP32" names="999" />
-			</output>
-		</layer>
-		<layer id="1234" name="__module.model.gpt_neox.layers.4.attention/aten::where/Select" type="Select" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="1" precision="FP32" />
-				<port id="2" precision="FP32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="1013,causal_mask.17">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1235" name="Add_290362" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1236" name="ShapeOf_290363" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1237" name="Constant_254389" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1238" name="__module.model.gpt_neox.layers.4.attention/aten::size/Constant_10" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1239" name="__module.model.gpt_neox.layers.4.attention/aten::size/Gather_10" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="997">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1240" name="__module.model.gpt_neox.layers.4.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1">
-			<data axis="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64" names="1014">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1241" name="__module.model.gpt_neox.layers.4.attention/aten::expand/Abs" type="Abs" version="opset1">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1242" name="Maximum_290364" type="Maximum" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1243" name="__module.model.gpt_neox.layers.4.attention/aten::add/Add_2" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1016">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1244" name="__module.model.gpt_neox.layers.4.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
-			<data causal="false" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1017,1018,sdpa_result.9,tensor.17">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1245" name="Constant_3016" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="1019">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1246" name="__module.model.gpt_neox.layers.4.attention/aten::permute/Transpose_3" type="Transpose" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1020">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1247" name="__module.model.gpt_neox.layers.4.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438650" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1248" name="__module.model.gpt_neox.layers.4.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
-			<data special_zero="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>3</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1025">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1249" name="self.model.gpt_neox.layers.4.attention.dense.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 2560" offset="520554194" size="6553600" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1250" name="Convert_21594" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1251" name="self.model.gpt_neox.layers.4.attention.dense.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="527107794" size="2560" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1252" name="Convert_21597" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1253" name="self.model.gpt_neox.layers.4.attention.dense.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1254" name="self.model.gpt_neox.layers.4.attention.dense.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="527110354" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1255" name="self.model.gpt_neox.layers.4.attention.dense.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1256" name="self.model.gpt_neox.layers.4.attention.dense.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1257" name="__module.model.gpt_neox.layers.4.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1258" name="Constant_299118_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="527115474" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1259" name="Constant_299118" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1260" name="__module.model.gpt_neox.layers.4.attention.dense/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1028,1030,input.19">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1261" name="__module.model.gpt_neox.layers.4/aten::add/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1048">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1262" name="__module.model.gpt_neox.layers.4/aten::add/Add_1" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1049,1051,hidden_states.11">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1263" name="__module.model.gpt_neox.layers.5.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867670" size="4" />
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1264" name="__module.model.gpt_neox.layers.5.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
-			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1265" name="Constant_299119_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="527120594" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1266" name="Constant_299119" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1267" name="__module.model.gpt_neox.layers.5.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1268" name="Constant_299120_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="527125714" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1269" name="Constant_299120" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1270" name="__module.model.gpt_neox.layers.5.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1186">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1271" name="self.model.gpt_neox.layers.5.mlp.dense_h_to_4h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 2560" offset="527130834" size="26214400" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1272" name="Convert_21644" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1273" name="self.model.gpt_neox.layers.5.mlp.dense_h_to_4h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 1" offset="553345234" size="10240" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1274" name="Convert_21647" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1275" name="self.model.gpt_neox.layers.5.mlp.dense_h_to_4h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1276" name="self.model.gpt_neox.layers.5.mlp.dense_h_to_4h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="10240, 1" offset="553355474" size="20480" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1277" name="self.model.gpt_neox.layers.5.mlp.dense_h_to_4h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1278" name="self.model.gpt_neox.layers.5.mlp.dense_h_to_4h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1279" name="__module.model.gpt_neox.layers.5.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1280" name="Constant_299121_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 10240" offset="553375954" size="20480" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1281" name="Constant_299121" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1282" name="__module.model.gpt_neox.layers.5.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1191">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1283" name="__module.model.gpt_neox.layers.5.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
-			<data approximation_mode="ERF" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32" names="1192">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1284" name="self.model.gpt_neox.layers.5.mlp.dense_4h_to_h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 10240" offset="553396434" size="26214400" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1285" name="Convert_21654" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1286" name="self.model.gpt_neox.layers.5.mlp.dense_4h_to_h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="579610834" size="2560" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1287" name="Convert_21657" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1288" name="self.model.gpt_neox.layers.5.mlp.dense_4h_to_h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1289" name="self.model.gpt_neox.layers.5.mlp.dense_4h_to_h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="579613394" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1290" name="self.model.gpt_neox.layers.5.mlp.dense_4h_to_h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1291" name="self.model.gpt_neox.layers.5.mlp.dense_4h_to_h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1292" name="__module.model.gpt_neox.layers.5.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1293" name="Constant_299122_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="579618514" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1294" name="Constant_299122" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1295" name="__module.model.gpt_neox.layers.5.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1195,input.25">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1296" name="Constant_299123_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="579623634" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1297" name="Constant_299123" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1298" name="__module.model.gpt_neox.layers.5.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1299" name="Constant_299124_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="579628754" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1300" name="Constant_299124" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1301" name="__module.model.gpt_neox.layers.5.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1061">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1302" name="self.model.gpt_neox.layers.5.attention.query_key_value.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 2560" offset="579633874" size="19660800" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1303" name="Convert_21624" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1304" name="self.model.gpt_neox.layers.5.attention.query_key_value.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 1" offset="599294674" size="7680" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1305" name="Convert_21627" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1306" name="self.model.gpt_neox.layers.5.attention.query_key_value.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1307" name="self.model.gpt_neox.layers.5.attention.query_key_value.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="7680, 1" offset="599302354" size="15360" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1308" name="self.model.gpt_neox.layers.5.attention.query_key_value.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1309" name="self.model.gpt_neox.layers.5.attention.query_key_value.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1310" name="__module.model.gpt_neox.layers.5.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1311" name="Constant_299125_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 7680" offset="599317714" size="15360" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1312" name="Constant_299125" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1313" name="__module.model.gpt_neox.layers.5.attention.query_key_value/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1068,qkv.21">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1314" name="__module.model.gpt_neox.layers.5.attention/prim::ListConstruct/Concat" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080154" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1315" name="__module.model.gpt_neox.layers.5.attention/aten::view/Reshape" type="Reshape" version="opset1">
-			<data special_zero="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1072,qkv.23">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1316" name="Constant_270904" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1317" name="Constant_270907" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1318" name="Constant_270910" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1319" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1073">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1320" name="Constant_3128" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="1074">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1321" name="__module.model.gpt_neox.layers.5.attention/aten::permute/Transpose" type="Transpose" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1075,query.31">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1322" name="Constant_270916" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1323" name="Constant_270919" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1324" name="Constant_270922" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1325" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1082,q.11">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1326" name="self.model.gpt_neox.layers.5.attention.rotary_emb.cos_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201080346" size="81920" />
-			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.5.attention.rotary_emb.cos_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1327" name="self.model.gpt_neox.layers.5.attention.rotary_emb.cos_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1328" name="__module.model.gpt_neox.layers.5.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1329" name="Constant_270928" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1330" name="Constant_270931" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1331" name="Constant_270934" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1332" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1076">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1333" name="Constant_3142" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="1077">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1334" name="__module.model.gpt_neox.layers.5.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1078,key.31">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1335" name="__module.model.gpt_neox.layers.5.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1336" name="__module.model.gpt_neox.layers.5.attention/aten::size/Constant_2" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1337" name="__module.model.gpt_neox.layers.5.attention/aten::size/Gather_2" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1086" />
-			</output>
-		</layer>
-		<layer id="1338" name="Constant_337559_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="1339" name="Constant_337559" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="1340" name="Constant_337552" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1341" name="Constant_337554" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1342" name="Constant_337556" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1343" name="Concat_337557" type="Concat" version="opset1">
-			<data axis="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1344" name="Broadcast_337560" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1345" name="ReadValue_330533" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.5.keypresent.5.key" variable_type="f32" variable_shape="?,32,?,80" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32" names="past_key_values.5.key">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1346" name="Constant_323415" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64" />
-			</output>
-		</layer>
-		<layer id="1347" name="Gather_323416" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I64" />
-			</input>
-			<output>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1348" name="__module.model.gpt_neox.layers.5.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1349" name="__module.model.gpt_neox.layers.5.attention/aten::size/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1350" name="__module.model.gpt_neox.layers.5.attention/aten::size/Gather_3" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1088" />
-			</output>
-		</layer>
-		<layer id="1351" name="__module.model.gpt_neox.layers.5.attention/aten::add_/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" names="1087,1091,1092,seq_len.21" />
-			</output>
-		</layer>
-		<layer id="1352" name="__module.model.gpt_neox.layers.5.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1353" name="__module.model.gpt_neox.layers.5.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1354" name="__module.model.gpt_neox.layers.5.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1355" name="__module.model.gpt_neox.layers.5.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1095,1098,cos.21">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1356" name="Constant_38227" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1357" name="__module.model.gpt_neox.layers.5.attention/aten::index/Gather" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="1101">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1358" name="__module.model.gpt_neox.layers.5.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1102,cos.23">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1359" name="__module.model.gpt_neox.layers.5.attention/aten::mul/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1106">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1360" name="Constant_270939" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1361" name="Constant_270938" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1362" name="__module.model.gpt_neox.layers.5.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1363" name="__module.model.gpt_neox.layers.5.attention/aten::size/Constant_5" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1364" name="__module.model.gpt_neox.layers.5.attention/aten::size/Gather_5" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1107,1112" />
-			</output>
-		</layer>
-		<layer id="1365" name="__module.model.gpt_neox.layers.5.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" />
-			</output>
-		</layer>
-		<layer id="1366" name="__module.model.gpt_neox.layers.5.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-			</input>
-			<output>
-				<port id="1" precision="I64" names="1109,1110,1114,1115" />
-			</output>
-		</layer>
-		<layer id="1367" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Constant_1" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1368" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1369" name="Constant_270937" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1370" name="ScatterUpdate_270940" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1371" name="Constant_270943" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1372" name="Constant_270946" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1373" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1116,x2.21">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1374" name="Constant_299126_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1375" name="Constant_299126" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1376" name="__module.model.gpt_neox.layers.5.attention/aten::neg/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1117">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1377" name="Constant_270991" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1378" name="Constant_270996" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1379" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1111">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1380" name="__module.model.gpt_neox.layers.5.attention/aten::cat/Concat" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1119">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1381" name="self.model.gpt_neox.layers.5.attention.rotary_emb.sin_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201162356" size="81920" />
-			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.5.attention.rotary_emb.sin_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1382" name="self.model.gpt_neox.layers.5.attention.rotary_emb.sin_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1383" name="__module.model.gpt_neox.layers.5.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1384" name="__module.model.gpt_neox.layers.5.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1385" name="__module.model.gpt_neox.layers.5.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1096,1099,sin.21">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1386" name="Constant_38230" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1387" name="__module.model.gpt_neox.layers.5.attention/aten::index/Gather_1" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="1104">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1388" name="__module.model.gpt_neox.layers.5.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1105,sin.23">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1389" name="__module.model.gpt_neox.layers.5.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1120">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1390" name="__module.model.gpt_neox.layers.5.attention/aten::add/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1121">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1391" name="Constant_271042" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1392" name="Constant_271045" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1393" name="Constant_271048" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1394" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1083">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1395" name="__module.model.gpt_neox.layers.5.attention/aten::cat/Concat_1" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1139,1149,query.33,query.35">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1396" name="Constant_271054" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1397" name="Constant_271057" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1398" name="Constant_271060" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1399" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1084,k.11">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1400" name="__module.model.gpt_neox.layers.5.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1122">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1401" name="Constant_271064" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1402" name="Constant_271063" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1403" name="__module.model.gpt_neox.layers.5.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1404" name="__module.model.gpt_neox.layers.5.attention/aten::size/Constant_7" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1405" name="__module.model.gpt_neox.layers.5.attention/aten::size/Gather_7" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1123,1128" />
-			</output>
-		</layer>
-		<layer id="1406" name="__module.model.gpt_neox.layers.5.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" />
-			</output>
-		</layer>
-		<layer id="1407" name="__module.model.gpt_neox.layers.5.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-			</input>
-			<output>
-				<port id="1" precision="I64" names="1125,1126,1130,1131" />
-			</output>
-		</layer>
-		<layer id="1408" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1409" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1410" name="Constant_271062" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1411" name="ScatterUpdate_271065" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1412" name="Constant_271068" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1413" name="Constant_271071" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1414" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1132,x2.23">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1415" name="Constant_299127_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1416" name="Constant_299127" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1417" name="__module.model.gpt_neox.layers.5.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1133">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1418" name="Constant_271116" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1419" name="Constant_271121" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1420" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1127">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1421" name="__module.model.gpt_neox.layers.5.attention/aten::cat/Concat_2" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1135">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1422" name="__module.model.gpt_neox.layers.5.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1136">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1423" name="__module.model.gpt_neox.layers.5.attention/aten::add/Add_1" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1137">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1424" name="Constant_271166" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1425" name="Constant_271169" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1426" name="Constant_271172" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1427" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1085">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1428" name="__module.model.gpt_neox.layers.5.attention/aten::cat/Concat_3" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1141">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1429" name="__module.model.gpt_neox.layers.5.attention/aten::to/Convert_1" type="Concat" version="opset1">
-			<data axis="-2" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="present.5.key">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1430" name="Constant_337572_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="1431" name="Constant_337572" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="1432" name="Constant_337565" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1433" name="Constant_337567" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1434" name="Constant_337569" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1435" name="Concat_337570" type="Concat" version="opset1">
-			<data axis="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1436" name="Broadcast_337573" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1437" name="ReadValue_330535" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.5.valuepresent.5.value" variable_type="f32" variable_shape="?,32,?,80" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32" names="past_key_values.5.value">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1438" name="Constant_323418" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64" />
-			</output>
-		</layer>
-		<layer id="1439" name="Gather_323419" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I64" />
-			</input>
-			<output>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1440" name="Constant_271178" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1441" name="Constant_271181" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1442" name="Constant_271184" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1443" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1079">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1444" name="Constant_3156" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="1080">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1445" name="__module.model.gpt_neox.layers.5.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1081">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1446" name="__module.model.gpt_neox.layers.5.attention/aten::cat/Concat_5" type="Concat" version="opset1">
-			<data axis="-2" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="present.5.value">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1447" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Slice_11" type="Const" version="opset1">
-			<data element_type="boolean" shape="1, 1, 2048, 2048" offset="201244276" size="4194304" />
-			<output>
-				<port id="0" precision="BOOL" names="1158">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1448" name="Constant_271188" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1449" name="Constant_271187" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162298" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1450" name="__module.model.gpt_neox.layers.5.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1451" name="__module.model.gpt_neox.layers.5.attention/aten::size/Constant_8" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1452" name="__module.model.gpt_neox.layers.5.attention/aten::size/Gather_8" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1153" />
-			</output>
-		</layer>
-		<layer id="1453" name="__module.model.gpt_neox.layers.5.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1454" name="__module.model.gpt_neox.layers.5.attention/aten::size/Constant_9" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1455" name="__module.model.gpt_neox.layers.5.attention/aten::size/Gather_9" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1151" />
-			</output>
-		</layer>
-		<layer id="1456" name="__module.model.gpt_neox.layers.5.attention/aten::sub/Subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" names="1155,1156" />
-			</output>
-		</layer>
-		<layer id="1457" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Constant_4" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1458" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1459" name="Constant_271186" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1460" name="ScatterUpdate_271189" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1461" name="Constant_271190" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1462" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1463" name="ScatterUpdate_271191" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1464" name="Constant_271194" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438612" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1465" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>3</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="BOOL" names="1159">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1466" name="Constant_271267" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1467" name="Constant_271268" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1468" name="Constant_271264" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1469" name="Constant_271263" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1470" name="ScatterUpdate_271269" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1471" name="Constant_271272" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1472" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="BOOL" names="1160,1161">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1473" name="__module.model.gpt_neox.layers.5.attention/aten::where/Convert_1_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="1474" name="__module.model.gpt_neox.layers.5.attention/aten::where/Convert_1" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="1475" name="__module.model.gpt_neox.layers.5.attention/aten::full/Convert" type="Const" version="opset1">
-			<data element_type="f32" shape="" offset="205438636" size="4" />
-			<output>
-				<port id="0" precision="FP32" names="1148" />
-			</output>
-		</layer>
-		<layer id="1476" name="__module.model.gpt_neox.layers.5.attention/aten::where/Select" type="Select" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="1" precision="FP32" />
-				<port id="2" precision="FP32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="1162,causal_mask.21">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1477" name="Add_290368" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1478" name="ShapeOf_290369" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1479" name="Constant_254404" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1480" name="__module.model.gpt_neox.layers.5.attention/aten::size/Constant_10" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1481" name="__module.model.gpt_neox.layers.5.attention/aten::size/Gather_10" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1146">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1482" name="__module.model.gpt_neox.layers.5.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1">
-			<data axis="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64" names="1163">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1483" name="__module.model.gpt_neox.layers.5.attention/aten::expand/Abs" type="Abs" version="opset1">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1484" name="Maximum_290370" type="Maximum" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1485" name="__module.model.gpt_neox.layers.5.attention/aten::add/Add_2" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1165">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1486" name="__module.model.gpt_neox.layers.5.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
-			<data causal="false" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1166,1167,sdpa_result.11,tensor.21">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1487" name="Constant_3597" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="1168">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1488" name="__module.model.gpt_neox.layers.5.attention/aten::permute/Transpose_3" type="Transpose" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1169">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1489" name="__module.model.gpt_neox.layers.5.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438650" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1490" name="__module.model.gpt_neox.layers.5.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
-			<data special_zero="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>3</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1174">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1491" name="self.model.gpt_neox.layers.5.attention.dense.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 2560" offset="599333074" size="6553600" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1492" name="Convert_21634" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1493" name="self.model.gpt_neox.layers.5.attention.dense.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="605886674" size="2560" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1494" name="Convert_21637" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1495" name="self.model.gpt_neox.layers.5.attention.dense.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1496" name="self.model.gpt_neox.layers.5.attention.dense.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="605889234" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1497" name="self.model.gpt_neox.layers.5.attention.dense.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1498" name="self.model.gpt_neox.layers.5.attention.dense.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1499" name="__module.model.gpt_neox.layers.5.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1500" name="Constant_299128_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="605894354" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1501" name="Constant_299128" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1502" name="__module.model.gpt_neox.layers.5.attention.dense/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1177,1179,input.23">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1503" name="__module.model.gpt_neox.layers.5/aten::add/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1197">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1504" name="__module.model.gpt_neox.layers.5/aten::add/Add_1" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1198,1200,hidden_states.13">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1505" name="__module.model.gpt_neox.layers.6.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867670" size="4" />
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1506" name="__module.model.gpt_neox.layers.6.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
-			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1507" name="Constant_299129_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="605899474" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1508" name="Constant_299129" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1509" name="__module.model.gpt_neox.layers.6.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1510" name="Constant_299130_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="605904594" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1511" name="Constant_299130" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1512" name="__module.model.gpt_neox.layers.6.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1335">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1513" name="self.model.gpt_neox.layers.6.mlp.dense_h_to_4h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 2560" offset="605909714" size="26214400" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1514" name="Convert_21684" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1515" name="self.model.gpt_neox.layers.6.mlp.dense_h_to_4h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 1" offset="632124114" size="10240" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1516" name="Convert_21687" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1517" name="self.model.gpt_neox.layers.6.mlp.dense_h_to_4h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1518" name="self.model.gpt_neox.layers.6.mlp.dense_h_to_4h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="10240, 1" offset="632134354" size="20480" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1519" name="self.model.gpt_neox.layers.6.mlp.dense_h_to_4h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1520" name="self.model.gpt_neox.layers.6.mlp.dense_h_to_4h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1521" name="__module.model.gpt_neox.layers.6.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1522" name="Constant_299131_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 10240" offset="632154834" size="20480" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1523" name="Constant_299131" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1524" name="__module.model.gpt_neox.layers.6.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1340">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1525" name="__module.model.gpt_neox.layers.6.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
-			<data approximation_mode="ERF" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32" names="1341">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1526" name="self.model.gpt_neox.layers.6.mlp.dense_4h_to_h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 10240" offset="632175314" size="26214400" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1527" name="Convert_21694" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1528" name="self.model.gpt_neox.layers.6.mlp.dense_4h_to_h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="658389714" size="2560" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1529" name="Convert_21697" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1530" name="self.model.gpt_neox.layers.6.mlp.dense_4h_to_h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1531" name="self.model.gpt_neox.layers.6.mlp.dense_4h_to_h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="658392274" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1532" name="self.model.gpt_neox.layers.6.mlp.dense_4h_to_h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1533" name="self.model.gpt_neox.layers.6.mlp.dense_4h_to_h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1534" name="__module.model.gpt_neox.layers.6.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1535" name="Constant_299132_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="658397394" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1536" name="Constant_299132" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1537" name="__module.model.gpt_neox.layers.6.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1344,input.29">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1538" name="Constant_299133_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="658402514" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1539" name="Constant_299133" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1540" name="__module.model.gpt_neox.layers.6.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1541" name="Constant_299134_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="658407634" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1542" name="Constant_299134" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1543" name="__module.model.gpt_neox.layers.6.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1210">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1544" name="self.model.gpt_neox.layers.6.attention.query_key_value.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 2560" offset="658412754" size="19660800" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1545" name="Convert_21664" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1546" name="self.model.gpt_neox.layers.6.attention.query_key_value.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 1" offset="678073554" size="7680" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1547" name="Convert_21667" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1548" name="self.model.gpt_neox.layers.6.attention.query_key_value.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1549" name="self.model.gpt_neox.layers.6.attention.query_key_value.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="7680, 1" offset="678081234" size="15360" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1550" name="self.model.gpt_neox.layers.6.attention.query_key_value.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1551" name="self.model.gpt_neox.layers.6.attention.query_key_value.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1552" name="__module.model.gpt_neox.layers.6.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1553" name="Constant_299135_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 7680" offset="678096594" size="15360" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1554" name="Constant_299135" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1555" name="__module.model.gpt_neox.layers.6.attention.query_key_value/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1217,qkv.25">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1556" name="__module.model.gpt_neox.layers.6.attention/prim::ListConstruct/Concat" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080154" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1557" name="__module.model.gpt_neox.layers.6.attention/aten::view/Reshape" type="Reshape" version="opset1">
-			<data special_zero="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1221,qkv.27">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1558" name="Constant_271279" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1559" name="Constant_271282" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1560" name="Constant_271285" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1561" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1222">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1562" name="Constant_3709" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="1223">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1563" name="__module.model.gpt_neox.layers.6.attention/aten::permute/Transpose" type="Transpose" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1224,query.37">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1564" name="Constant_271291" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1565" name="Constant_271294" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1566" name="Constant_271297" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1567" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1231,q.13">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1568" name="self.model.gpt_neox.layers.6.attention.rotary_emb.cos_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201080346" size="81920" />
-			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.6.attention.rotary_emb.cos_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1569" name="self.model.gpt_neox.layers.6.attention.rotary_emb.cos_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1570" name="__module.model.gpt_neox.layers.6.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1571" name="Constant_271303" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1572" name="Constant_271306" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1573" name="Constant_271309" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1574" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1225">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1575" name="Constant_3723" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="1226">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1576" name="__module.model.gpt_neox.layers.6.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1227,key.37">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1577" name="__module.model.gpt_neox.layers.6.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1578" name="__module.model.gpt_neox.layers.6.attention/aten::size/Constant_2" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1579" name="__module.model.gpt_neox.layers.6.attention/aten::size/Gather_2" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1235" />
-			</output>
-		</layer>
-		<layer id="1580" name="Constant_337585_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="1581" name="Constant_337585" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="1582" name="Constant_337578" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1583" name="Constant_337580" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1584" name="Constant_337582" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1585" name="Concat_337583" type="Concat" version="opset1">
-			<data axis="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1586" name="Broadcast_337586" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1587" name="ReadValue_330537" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.6.keypresent.6.key" variable_type="f32" variable_shape="?,32,?,80" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32" names="past_key_values.6.key">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1588" name="Constant_323421" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64" />
-			</output>
-		</layer>
-		<layer id="1589" name="Gather_323422" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I64" />
-			</input>
-			<output>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1590" name="__module.model.gpt_neox.layers.6.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1591" name="__module.model.gpt_neox.layers.6.attention/aten::size/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1592" name="__module.model.gpt_neox.layers.6.attention/aten::size/Gather_3" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1237" />
-			</output>
-		</layer>
-		<layer id="1593" name="__module.model.gpt_neox.layers.6.attention/aten::add_/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" names="1236,1240,1241,seq_len.25" />
-			</output>
-		</layer>
-		<layer id="1594" name="__module.model.gpt_neox.layers.6.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1595" name="__module.model.gpt_neox.layers.6.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1596" name="__module.model.gpt_neox.layers.6.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1597" name="__module.model.gpt_neox.layers.6.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1244,1247,cos.25">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1598" name="Constant_38233" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1599" name="__module.model.gpt_neox.layers.6.attention/aten::index/Gather" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="1250">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1600" name="__module.model.gpt_neox.layers.6.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1251,cos.27">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1601" name="__module.model.gpt_neox.layers.6.attention/aten::mul/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1255">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1602" name="Constant_271314" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1603" name="Constant_271313" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1604" name="__module.model.gpt_neox.layers.6.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1605" name="__module.model.gpt_neox.layers.6.attention/aten::size/Constant_5" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1606" name="__module.model.gpt_neox.layers.6.attention/aten::size/Gather_5" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1256,1261" />
-			</output>
-		</layer>
-		<layer id="1607" name="__module.model.gpt_neox.layers.6.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" />
-			</output>
-		</layer>
-		<layer id="1608" name="__module.model.gpt_neox.layers.6.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-			</input>
-			<output>
-				<port id="1" precision="I64" names="1258,1259,1263,1264" />
-			</output>
-		</layer>
-		<layer id="1609" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Constant_1" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1610" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1611" name="Constant_271312" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1612" name="ScatterUpdate_271315" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1613" name="Constant_271318" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1614" name="Constant_271321" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1615" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1265,x2.25">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1616" name="Constant_299136_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1617" name="Constant_299136" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1618" name="__module.model.gpt_neox.layers.6.attention/aten::neg/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1266">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1619" name="Constant_271366" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1620" name="Constant_271371" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1621" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1260">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1622" name="__module.model.gpt_neox.layers.6.attention/aten::cat/Concat" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1268">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1623" name="self.model.gpt_neox.layers.6.attention.rotary_emb.sin_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201162356" size="81920" />
-			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.6.attention.rotary_emb.sin_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1624" name="self.model.gpt_neox.layers.6.attention.rotary_emb.sin_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1625" name="__module.model.gpt_neox.layers.6.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1626" name="__module.model.gpt_neox.layers.6.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1627" name="__module.model.gpt_neox.layers.6.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1245,1248,sin.25">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1628" name="Constant_38236" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1629" name="__module.model.gpt_neox.layers.6.attention/aten::index/Gather_1" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="1253">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1630" name="__module.model.gpt_neox.layers.6.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1254,sin.27">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1631" name="__module.model.gpt_neox.layers.6.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1269">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1632" name="__module.model.gpt_neox.layers.6.attention/aten::add/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1270">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1633" name="Constant_271417" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1634" name="Constant_271420" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1635" name="Constant_271423" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1636" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1232">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1637" name="__module.model.gpt_neox.layers.6.attention/aten::cat/Concat_1" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1288,1298,query.39,query.41">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1638" name="Constant_271429" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1639" name="Constant_271432" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1640" name="Constant_271435" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1641" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1233,k.13">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1642" name="__module.model.gpt_neox.layers.6.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1271">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1643" name="Constant_271439" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1644" name="Constant_271438" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1645" name="__module.model.gpt_neox.layers.6.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1646" name="__module.model.gpt_neox.layers.6.attention/aten::size/Constant_7" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1647" name="__module.model.gpt_neox.layers.6.attention/aten::size/Gather_7" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1272,1277" />
-			</output>
-		</layer>
-		<layer id="1648" name="__module.model.gpt_neox.layers.6.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" />
-			</output>
-		</layer>
-		<layer id="1649" name="__module.model.gpt_neox.layers.6.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-			</input>
-			<output>
-				<port id="1" precision="I64" names="1274,1275,1279,1280" />
-			</output>
-		</layer>
-		<layer id="1650" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1651" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1652" name="Constant_271437" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1653" name="ScatterUpdate_271440" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1654" name="Constant_271443" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1655" name="Constant_271446" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1656" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1281,x2.27">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1657" name="Constant_299137_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1658" name="Constant_299137" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1659" name="__module.model.gpt_neox.layers.6.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1282">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1660" name="Constant_271491" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1661" name="Constant_271496" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1662" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1276">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1663" name="__module.model.gpt_neox.layers.6.attention/aten::cat/Concat_2" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1284">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1664" name="__module.model.gpt_neox.layers.6.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1285">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1665" name="__module.model.gpt_neox.layers.6.attention/aten::add/Add_1" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1286">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1666" name="Constant_271541" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1667" name="Constant_271544" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1668" name="Constant_271547" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1669" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1234">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1670" name="__module.model.gpt_neox.layers.6.attention/aten::cat/Concat_3" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1290">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1671" name="__module.model.gpt_neox.layers.6.attention/aten::to/Convert_1" type="Concat" version="opset1">
-			<data axis="-2" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="present.6.key">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1672" name="Constant_337598_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="1673" name="Constant_337598" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="1674" name="Constant_337591" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1675" name="Constant_337593" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1676" name="Constant_337595" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1677" name="Concat_337596" type="Concat" version="opset1">
-			<data axis="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1678" name="Broadcast_337599" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1679" name="ReadValue_330539" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.6.valuepresent.6.value" variable_type="f32" variable_shape="?,32,?,80" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32" names="past_key_values.6.value">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1680" name="Constant_323424" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64" />
-			</output>
-		</layer>
-		<layer id="1681" name="Gather_323425" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I64" />
-			</input>
-			<output>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1682" name="Constant_271553" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1683" name="Constant_271556" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1684" name="Constant_271559" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1685" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1228">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1686" name="Constant_3737" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="1229">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1687" name="__module.model.gpt_neox.layers.6.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1230">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1688" name="__module.model.gpt_neox.layers.6.attention/aten::cat/Concat_5" type="Concat" version="opset1">
-			<data axis="-2" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="present.6.value">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1689" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Slice_11" type="Const" version="opset1">
-			<data element_type="boolean" shape="1, 1, 2048, 2048" offset="201244276" size="4194304" />
-			<output>
-				<port id="0" precision="BOOL" names="1307">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1690" name="Constant_271563" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1691" name="Constant_271562" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162298" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1692" name="__module.model.gpt_neox.layers.6.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1693" name="__module.model.gpt_neox.layers.6.attention/aten::size/Constant_8" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1694" name="__module.model.gpt_neox.layers.6.attention/aten::size/Gather_8" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1302" />
-			</output>
-		</layer>
-		<layer id="1695" name="__module.model.gpt_neox.layers.6.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1696" name="__module.model.gpt_neox.layers.6.attention/aten::size/Constant_9" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1697" name="__module.model.gpt_neox.layers.6.attention/aten::size/Gather_9" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1300" />
-			</output>
-		</layer>
-		<layer id="1698" name="__module.model.gpt_neox.layers.6.attention/aten::sub/Subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" names="1304,1305" />
-			</output>
-		</layer>
-		<layer id="1699" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Constant_4" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1700" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1701" name="Constant_271561" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1702" name="ScatterUpdate_271564" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1703" name="Constant_271565" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1704" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1705" name="ScatterUpdate_271566" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1706" name="Constant_271569" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438612" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1707" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>3</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="BOOL" names="1308">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1708" name="Constant_271642" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1709" name="Constant_271643" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1710" name="Constant_271639" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1711" name="Constant_271638" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1712" name="ScatterUpdate_271644" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1713" name="Constant_271647" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1714" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="BOOL" names="1309,1310">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1715" name="__module.model.gpt_neox.layers.6.attention/aten::where/Convert_1_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="1716" name="__module.model.gpt_neox.layers.6.attention/aten::where/Convert_1" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="1717" name="__module.model.gpt_neox.layers.6.attention/aten::full/Convert" type="Const" version="opset1">
-			<data element_type="f32" shape="" offset="205438636" size="4" />
-			<output>
-				<port id="0" precision="FP32" names="1297" />
-			</output>
-		</layer>
-		<layer id="1718" name="__module.model.gpt_neox.layers.6.attention/aten::where/Select" type="Select" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="1" precision="FP32" />
-				<port id="2" precision="FP32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="1311,causal_mask.25">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1719" name="Add_290374" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1720" name="ShapeOf_290375" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1721" name="Constant_254419" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1722" name="__module.model.gpt_neox.layers.6.attention/aten::size/Constant_10" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1723" name="__module.model.gpt_neox.layers.6.attention/aten::size/Gather_10" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1295">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1724" name="__module.model.gpt_neox.layers.6.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1">
-			<data axis="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64" names="1312">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1725" name="__module.model.gpt_neox.layers.6.attention/aten::expand/Abs" type="Abs" version="opset1">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1726" name="Maximum_290376" type="Maximum" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1727" name="__module.model.gpt_neox.layers.6.attention/aten::add/Add_2" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1314">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1728" name="__module.model.gpt_neox.layers.6.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
-			<data causal="false" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1315,1316,sdpa_result.13,tensor.25">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1729" name="Constant_4178" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="1317">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1730" name="__module.model.gpt_neox.layers.6.attention/aten::permute/Transpose_3" type="Transpose" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1318">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1731" name="__module.model.gpt_neox.layers.6.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438650" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1732" name="__module.model.gpt_neox.layers.6.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
-			<data special_zero="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>3</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1323">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1733" name="self.model.gpt_neox.layers.6.attention.dense.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 2560" offset="678111954" size="6553600" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1734" name="Convert_21674" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1735" name="self.model.gpt_neox.layers.6.attention.dense.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="684665554" size="2560" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1736" name="Convert_21677" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1737" name="self.model.gpt_neox.layers.6.attention.dense.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1738" name="self.model.gpt_neox.layers.6.attention.dense.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="684668114" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1739" name="self.model.gpt_neox.layers.6.attention.dense.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1740" name="self.model.gpt_neox.layers.6.attention.dense.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1741" name="__module.model.gpt_neox.layers.6.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1742" name="Constant_299138_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="684673234" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1743" name="Constant_299138" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1744" name="__module.model.gpt_neox.layers.6.attention.dense/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1326,1328,input.27">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1745" name="__module.model.gpt_neox.layers.6/aten::add/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1346">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1746" name="__module.model.gpt_neox.layers.6/aten::add/Add_1" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1347,1349,hidden_states.15">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1747" name="__module.model.gpt_neox.layers.7.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867670" size="4" />
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1748" name="__module.model.gpt_neox.layers.7.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
-			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1749" name="Constant_299139_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="684678354" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1750" name="Constant_299139" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1751" name="__module.model.gpt_neox.layers.7.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1752" name="Constant_299140_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="684683474" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1753" name="Constant_299140" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1754" name="__module.model.gpt_neox.layers.7.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1484">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1755" name="self.model.gpt_neox.layers.7.mlp.dense_h_to_4h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 2560" offset="684688594" size="26214400" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1756" name="Convert_21724" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1757" name="self.model.gpt_neox.layers.7.mlp.dense_h_to_4h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 1" offset="710902994" size="10240" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1758" name="Convert_21727" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1759" name="self.model.gpt_neox.layers.7.mlp.dense_h_to_4h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1760" name="self.model.gpt_neox.layers.7.mlp.dense_h_to_4h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="10240, 1" offset="710913234" size="20480" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1761" name="self.model.gpt_neox.layers.7.mlp.dense_h_to_4h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1762" name="self.model.gpt_neox.layers.7.mlp.dense_h_to_4h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1763" name="__module.model.gpt_neox.layers.7.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1764" name="Constant_299141_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 10240" offset="710933714" size="20480" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1765" name="Constant_299141" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1766" name="__module.model.gpt_neox.layers.7.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1489">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1767" name="__module.model.gpt_neox.layers.7.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
-			<data approximation_mode="ERF" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32" names="1490">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1768" name="self.model.gpt_neox.layers.7.mlp.dense_4h_to_h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 10240" offset="710954194" size="26214400" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1769" name="Convert_21734" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1770" name="self.model.gpt_neox.layers.7.mlp.dense_4h_to_h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="737168594" size="2560" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1771" name="Convert_21737" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1772" name="self.model.gpt_neox.layers.7.mlp.dense_4h_to_h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1773" name="self.model.gpt_neox.layers.7.mlp.dense_4h_to_h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="737171154" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1774" name="self.model.gpt_neox.layers.7.mlp.dense_4h_to_h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1775" name="self.model.gpt_neox.layers.7.mlp.dense_4h_to_h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1776" name="__module.model.gpt_neox.layers.7.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1777" name="Constant_299142_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="737176274" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1778" name="Constant_299142" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1779" name="__module.model.gpt_neox.layers.7.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1493,input.33">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1780" name="Constant_299143_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="737181394" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1781" name="Constant_299143" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1782" name="__module.model.gpt_neox.layers.7.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1783" name="Constant_299144_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="737186514" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1784" name="Constant_299144" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1785" name="__module.model.gpt_neox.layers.7.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1359">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1786" name="self.model.gpt_neox.layers.7.attention.query_key_value.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 2560" offset="737191634" size="19660800" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1787" name="Convert_21704" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1788" name="self.model.gpt_neox.layers.7.attention.query_key_value.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 1" offset="756852434" size="7680" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1789" name="Convert_21707" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1790" name="self.model.gpt_neox.layers.7.attention.query_key_value.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1791" name="self.model.gpt_neox.layers.7.attention.query_key_value.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="7680, 1" offset="756860114" size="15360" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1792" name="self.model.gpt_neox.layers.7.attention.query_key_value.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1793" name="self.model.gpt_neox.layers.7.attention.query_key_value.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1794" name="__module.model.gpt_neox.layers.7.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1795" name="Constant_299145_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 7680" offset="756875474" size="15360" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1796" name="Constant_299145" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1797" name="__module.model.gpt_neox.layers.7.attention.query_key_value/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1366,qkv.29">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1798" name="__module.model.gpt_neox.layers.7.attention/prim::ListConstruct/Concat" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080154" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1799" name="__module.model.gpt_neox.layers.7.attention/aten::view/Reshape" type="Reshape" version="opset1">
-			<data special_zero="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1370,qkv.31">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1800" name="Constant_271654" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1801" name="Constant_271657" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1802" name="Constant_271660" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1803" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1371">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1804" name="Constant_4290" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="1372">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1805" name="__module.model.gpt_neox.layers.7.attention/aten::permute/Transpose" type="Transpose" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1373,query.43">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1806" name="Constant_271666" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1807" name="Constant_271669" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1808" name="Constant_271672" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1809" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1380,q.15">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1810" name="self.model.gpt_neox.layers.7.attention.rotary_emb.cos_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201080346" size="81920" />
-			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.7.attention.rotary_emb.cos_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1811" name="self.model.gpt_neox.layers.7.attention.rotary_emb.cos_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1812" name="__module.model.gpt_neox.layers.7.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1813" name="Constant_271678" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1814" name="Constant_271681" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1815" name="Constant_271684" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1816" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1374">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1817" name="Constant_4304" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="1375">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1818" name="__module.model.gpt_neox.layers.7.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1376,key.43">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1819" name="__module.model.gpt_neox.layers.7.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1820" name="__module.model.gpt_neox.layers.7.attention/aten::size/Constant_2" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1821" name="__module.model.gpt_neox.layers.7.attention/aten::size/Gather_2" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1384" />
-			</output>
-		</layer>
-		<layer id="1822" name="Constant_337611_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="1823" name="Constant_337611" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="1824" name="Constant_337604" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1825" name="Constant_337606" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1826" name="Constant_337608" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1827" name="Concat_337609" type="Concat" version="opset1">
-			<data axis="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1828" name="Broadcast_337612" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1829" name="ReadValue_330541" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.7.keypresent.7.key" variable_type="f32" variable_shape="?,32,?,80" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32" names="past_key_values.7.key">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1830" name="Constant_323427" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64" />
-			</output>
-		</layer>
-		<layer id="1831" name="Gather_323428" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I64" />
-			</input>
-			<output>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1832" name="__module.model.gpt_neox.layers.7.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1833" name="__module.model.gpt_neox.layers.7.attention/aten::size/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1834" name="__module.model.gpt_neox.layers.7.attention/aten::size/Gather_3" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1386" />
-			</output>
-		</layer>
-		<layer id="1835" name="__module.model.gpt_neox.layers.7.attention/aten::add_/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" names="1385,1389,1390,seq_len.29" />
-			</output>
-		</layer>
-		<layer id="1836" name="__module.model.gpt_neox.layers.7.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1837" name="__module.model.gpt_neox.layers.7.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1838" name="__module.model.gpt_neox.layers.7.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1839" name="__module.model.gpt_neox.layers.7.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1393,1396,cos.29">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1840" name="Constant_38239" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1841" name="__module.model.gpt_neox.layers.7.attention/aten::index/Gather" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="1399">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1842" name="__module.model.gpt_neox.layers.7.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1400,cos.31">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1843" name="__module.model.gpt_neox.layers.7.attention/aten::mul/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1404">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1844" name="Constant_271689" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1845" name="Constant_271688" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1846" name="__module.model.gpt_neox.layers.7.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1847" name="__module.model.gpt_neox.layers.7.attention/aten::size/Constant_5" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1848" name="__module.model.gpt_neox.layers.7.attention/aten::size/Gather_5" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1405,1410" />
-			</output>
-		</layer>
-		<layer id="1849" name="__module.model.gpt_neox.layers.7.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" />
-			</output>
-		</layer>
-		<layer id="1850" name="__module.model.gpt_neox.layers.7.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-			</input>
-			<output>
-				<port id="1" precision="I64" names="1407,1408,1412,1413" />
-			</output>
-		</layer>
-		<layer id="1851" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Constant_1" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1852" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1853" name="Constant_271687" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1854" name="ScatterUpdate_271690" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1855" name="Constant_271693" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1856" name="Constant_271696" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1857" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1414,x2.29">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1858" name="Constant_299146_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1859" name="Constant_299146" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1860" name="__module.model.gpt_neox.layers.7.attention/aten::neg/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1415">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1861" name="Constant_271741" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1862" name="Constant_271746" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1863" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1409">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1864" name="__module.model.gpt_neox.layers.7.attention/aten::cat/Concat" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1417">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1865" name="self.model.gpt_neox.layers.7.attention.rotary_emb.sin_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201162356" size="81920" />
-			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.7.attention.rotary_emb.sin_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1866" name="self.model.gpt_neox.layers.7.attention.rotary_emb.sin_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1867" name="__module.model.gpt_neox.layers.7.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1868" name="__module.model.gpt_neox.layers.7.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1869" name="__module.model.gpt_neox.layers.7.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1394,1397,sin.29">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1870" name="Constant_38242" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1871" name="__module.model.gpt_neox.layers.7.attention/aten::index/Gather_1" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="1402">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1872" name="__module.model.gpt_neox.layers.7.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1403,sin.31">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1873" name="__module.model.gpt_neox.layers.7.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1418">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1874" name="__module.model.gpt_neox.layers.7.attention/aten::add/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1419">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1875" name="Constant_271792" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1876" name="Constant_271795" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1877" name="Constant_271798" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1878" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1381">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1879" name="__module.model.gpt_neox.layers.7.attention/aten::cat/Concat_1" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1437,1447,query.45,query.47">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1880" name="Constant_271804" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1881" name="Constant_271807" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1882" name="Constant_271810" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1883" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1382,k.15">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1884" name="__module.model.gpt_neox.layers.7.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1420">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1885" name="Constant_271814" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1886" name="Constant_271813" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1887" name="__module.model.gpt_neox.layers.7.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1888" name="__module.model.gpt_neox.layers.7.attention/aten::size/Constant_7" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1889" name="__module.model.gpt_neox.layers.7.attention/aten::size/Gather_7" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1421,1426" />
-			</output>
-		</layer>
-		<layer id="1890" name="__module.model.gpt_neox.layers.7.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" />
-			</output>
-		</layer>
-		<layer id="1891" name="__module.model.gpt_neox.layers.7.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-			</input>
-			<output>
-				<port id="1" precision="I64" names="1423,1424,1428,1429" />
-			</output>
-		</layer>
-		<layer id="1892" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1893" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1894" name="Constant_271812" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1895" name="ScatterUpdate_271815" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1896" name="Constant_271818" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1897" name="Constant_271821" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1898" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1430,x2.31">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1899" name="Constant_299147_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1900" name="Constant_299147" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1901" name="__module.model.gpt_neox.layers.7.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1431">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1902" name="Constant_271866" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1903" name="Constant_271871" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1904" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1425">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1905" name="__module.model.gpt_neox.layers.7.attention/aten::cat/Concat_2" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1433">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1906" name="__module.model.gpt_neox.layers.7.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1434">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1907" name="__module.model.gpt_neox.layers.7.attention/aten::add/Add_1" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1435">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1908" name="Constant_271916" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1909" name="Constant_271919" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1910" name="Constant_271922" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1911" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1383">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1912" name="__module.model.gpt_neox.layers.7.attention/aten::cat/Concat_3" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1439">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1913" name="__module.model.gpt_neox.layers.7.attention/aten::to/Convert_1" type="Concat" version="opset1">
-			<data axis="-2" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="present.7.key">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1914" name="Constant_337624_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="1915" name="Constant_337624" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="1916" name="Constant_337617" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1917" name="Constant_337619" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1918" name="Constant_337621" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1919" name="Concat_337622" type="Concat" version="opset1">
-			<data axis="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1920" name="Broadcast_337625" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1921" name="ReadValue_330543" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.7.valuepresent.7.value" variable_type="f32" variable_shape="?,32,?,80" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32" names="past_key_values.7.value">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1922" name="Constant_323430" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64" />
-			</output>
-		</layer>
-		<layer id="1923" name="Gather_323431" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I64" />
-			</input>
-			<output>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1924" name="Constant_271928" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1925" name="Constant_271931" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1926" name="Constant_271934" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1927" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1377">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1928" name="Constant_4318" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="1378">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1929" name="__module.model.gpt_neox.layers.7.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1379">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1930" name="__module.model.gpt_neox.layers.7.attention/aten::cat/Concat_5" type="Concat" version="opset1">
-			<data axis="-2" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="present.7.value">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1931" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Slice_11" type="Const" version="opset1">
-			<data element_type="boolean" shape="1, 1, 2048, 2048" offset="201244276" size="4194304" />
-			<output>
-				<port id="0" precision="BOOL" names="1456">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1932" name="Constant_271938" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1933" name="Constant_271937" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162298" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1934" name="__module.model.gpt_neox.layers.7.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1935" name="__module.model.gpt_neox.layers.7.attention/aten::size/Constant_8" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1936" name="__module.model.gpt_neox.layers.7.attention/aten::size/Gather_8" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1451" />
-			</output>
-		</layer>
-		<layer id="1937" name="__module.model.gpt_neox.layers.7.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1938" name="__module.model.gpt_neox.layers.7.attention/aten::size/Constant_9" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1939" name="__module.model.gpt_neox.layers.7.attention/aten::size/Gather_9" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1449" />
-			</output>
-		</layer>
-		<layer id="1940" name="__module.model.gpt_neox.layers.7.attention/aten::sub/Subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" names="1453,1454" />
-			</output>
-		</layer>
-		<layer id="1941" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Constant_4" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1942" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1943" name="Constant_271936" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1944" name="ScatterUpdate_271939" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1945" name="Constant_271940" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1946" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1947" name="ScatterUpdate_271941" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1948" name="Constant_271944" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438612" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1949" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>3</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="BOOL" names="1457">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1950" name="Constant_272017" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1951" name="Constant_272018" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1952" name="Constant_272014" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1953" name="Constant_272013" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1954" name="ScatterUpdate_272019" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1955" name="Constant_272022" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1956" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="BOOL" names="1458,1459">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1957" name="__module.model.gpt_neox.layers.7.attention/aten::where/Convert_1_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="1958" name="__module.model.gpt_neox.layers.7.attention/aten::where/Convert_1" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="1959" name="__module.model.gpt_neox.layers.7.attention/aten::full/Convert" type="Const" version="opset1">
-			<data element_type="f32" shape="" offset="205438636" size="4" />
-			<output>
-				<port id="0" precision="FP32" names="1446" />
-			</output>
-		</layer>
-		<layer id="1960" name="__module.model.gpt_neox.layers.7.attention/aten::where/Select" type="Select" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="1" precision="FP32" />
-				<port id="2" precision="FP32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="1460,causal_mask.29">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1961" name="Add_290380" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1962" name="ShapeOf_290381" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1963" name="Constant_254434" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1964" name="__module.model.gpt_neox.layers.7.attention/aten::size/Constant_10" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="1965" name="__module.model.gpt_neox.layers.7.attention/aten::size/Gather_10" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1444">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1966" name="__module.model.gpt_neox.layers.7.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1">
-			<data axis="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64" names="1461">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1967" name="__module.model.gpt_neox.layers.7.attention/aten::expand/Abs" type="Abs" version="opset1">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1968" name="Maximum_290382" type="Maximum" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1969" name="__module.model.gpt_neox.layers.7.attention/aten::add/Add_2" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1463">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1970" name="__module.model.gpt_neox.layers.7.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
-			<data causal="false" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1464,1465,sdpa_result.15,tensor.29">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1971" name="Constant_4759" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="1466">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1972" name="__module.model.gpt_neox.layers.7.attention/aten::permute/Transpose_3" type="Transpose" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1467">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1973" name="__module.model.gpt_neox.layers.7.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438650" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1974" name="__module.model.gpt_neox.layers.7.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
-			<data special_zero="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>3</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1472">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1975" name="self.model.gpt_neox.layers.7.attention.dense.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 2560" offset="756890834" size="6553600" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1976" name="Convert_21714" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1977" name="self.model.gpt_neox.layers.7.attention.dense.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="763444434" size="2560" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1978" name="Convert_21717" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1979" name="self.model.gpt_neox.layers.7.attention.dense.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1980" name="self.model.gpt_neox.layers.7.attention.dense.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="763446994" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1981" name="self.model.gpt_neox.layers.7.attention.dense.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1982" name="self.model.gpt_neox.layers.7.attention.dense.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1983" name="__module.model.gpt_neox.layers.7.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1984" name="Constant_299148_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="763452114" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1985" name="Constant_299148" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1986" name="__module.model.gpt_neox.layers.7.attention.dense/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1475,1477,input.31">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1987" name="__module.model.gpt_neox.layers.7/aten::add/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1495">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1988" name="__module.model.gpt_neox.layers.7/aten::add/Add_1" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1496,1498,hidden_states.17">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1989" name="__module.model.gpt_neox.layers.8.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867670" size="4" />
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1990" name="__module.model.gpt_neox.layers.8.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
-			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1991" name="Constant_299149_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="763457234" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1992" name="Constant_299149" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1993" name="__module.model.gpt_neox.layers.8.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1994" name="Constant_299150_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="763462354" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1995" name="Constant_299150" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1996" name="__module.model.gpt_neox.layers.8.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1633">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1997" name="self.model.gpt_neox.layers.8.mlp.dense_h_to_4h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 2560" offset="763467474" size="26214400" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1998" name="Convert_21764" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="1999" name="self.model.gpt_neox.layers.8.mlp.dense_h_to_4h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 1" offset="789681874" size="10240" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2000" name="Convert_21767" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2001" name="self.model.gpt_neox.layers.8.mlp.dense_h_to_4h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2002" name="self.model.gpt_neox.layers.8.mlp.dense_h_to_4h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="10240, 1" offset="789692114" size="20480" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2003" name="self.model.gpt_neox.layers.8.mlp.dense_h_to_4h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2004" name="self.model.gpt_neox.layers.8.mlp.dense_h_to_4h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2005" name="__module.model.gpt_neox.layers.8.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2006" name="Constant_299151_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 10240" offset="789712594" size="20480" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2007" name="Constant_299151" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2008" name="__module.model.gpt_neox.layers.8.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1638">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2009" name="__module.model.gpt_neox.layers.8.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
-			<data approximation_mode="ERF" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32" names="1639">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2010" name="self.model.gpt_neox.layers.8.mlp.dense_4h_to_h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 10240" offset="789733074" size="26214400" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2011" name="Convert_21774" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2012" name="self.model.gpt_neox.layers.8.mlp.dense_4h_to_h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="815947474" size="2560" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2013" name="Convert_21777" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2014" name="self.model.gpt_neox.layers.8.mlp.dense_4h_to_h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2015" name="self.model.gpt_neox.layers.8.mlp.dense_4h_to_h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="815950034" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2016" name="self.model.gpt_neox.layers.8.mlp.dense_4h_to_h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2017" name="self.model.gpt_neox.layers.8.mlp.dense_4h_to_h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2018" name="__module.model.gpt_neox.layers.8.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2019" name="Constant_299152_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="815955154" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2020" name="Constant_299152" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2021" name="__module.model.gpt_neox.layers.8.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1642,input.37">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2022" name="Constant_299153_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="815960274" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2023" name="Constant_299153" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2024" name="__module.model.gpt_neox.layers.8.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2025" name="Constant_299154_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="815965394" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2026" name="Constant_299154" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2027" name="__module.model.gpt_neox.layers.8.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1508">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2028" name="self.model.gpt_neox.layers.8.attention.query_key_value.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 2560" offset="815970514" size="19660800" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2029" name="Convert_21744" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2030" name="self.model.gpt_neox.layers.8.attention.query_key_value.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 1" offset="835631314" size="7680" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2031" name="Convert_21747" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2032" name="self.model.gpt_neox.layers.8.attention.query_key_value.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2033" name="self.model.gpt_neox.layers.8.attention.query_key_value.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="7680, 1" offset="835638994" size="15360" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2034" name="self.model.gpt_neox.layers.8.attention.query_key_value.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2035" name="self.model.gpt_neox.layers.8.attention.query_key_value.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2036" name="__module.model.gpt_neox.layers.8.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2037" name="Constant_299155_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 7680" offset="835654354" size="15360" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2038" name="Constant_299155" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2039" name="__module.model.gpt_neox.layers.8.attention.query_key_value/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1515,qkv.33">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2040" name="__module.model.gpt_neox.layers.8.attention/prim::ListConstruct/Concat" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080154" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2041" name="__module.model.gpt_neox.layers.8.attention/aten::view/Reshape" type="Reshape" version="opset1">
-			<data special_zero="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1519,qkv.35">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2042" name="Constant_272029" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2043" name="Constant_272032" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2044" name="Constant_272035" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2045" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1520">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2046" name="Constant_4871" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="1521">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2047" name="__module.model.gpt_neox.layers.8.attention/aten::permute/Transpose" type="Transpose" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1522,query.49">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2048" name="Constant_272041" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2049" name="Constant_272044" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2050" name="Constant_272047" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2051" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1529,q.17">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2052" name="self.model.gpt_neox.layers.8.attention.rotary_emb.cos_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201080346" size="81920" />
-			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.8.attention.rotary_emb.cos_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2053" name="self.model.gpt_neox.layers.8.attention.rotary_emb.cos_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2054" name="__module.model.gpt_neox.layers.8.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2055" name="Constant_272053" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2056" name="Constant_272056" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2057" name="Constant_272059" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2058" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1523">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2059" name="Constant_4885" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="1524">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2060" name="__module.model.gpt_neox.layers.8.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1525,key.49">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2061" name="__module.model.gpt_neox.layers.8.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2062" name="__module.model.gpt_neox.layers.8.attention/aten::size/Constant_2" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2063" name="__module.model.gpt_neox.layers.8.attention/aten::size/Gather_2" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1533" />
-			</output>
-		</layer>
-		<layer id="2064" name="Constant_337637_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="2065" name="Constant_337637" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="2066" name="Constant_337630" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2067" name="Constant_337632" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2068" name="Constant_337634" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2069" name="Concat_337635" type="Concat" version="opset1">
-			<data axis="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2070" name="Broadcast_337638" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2071" name="ReadValue_330545" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.8.keypresent.8.key" variable_type="f32" variable_shape="?,32,?,80" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32" names="past_key_values.8.key">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2072" name="Constant_323433" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64" />
-			</output>
-		</layer>
-		<layer id="2073" name="Gather_323434" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I64" />
-			</input>
-			<output>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2074" name="__module.model.gpt_neox.layers.8.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2075" name="__module.model.gpt_neox.layers.8.attention/aten::size/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2076" name="__module.model.gpt_neox.layers.8.attention/aten::size/Gather_3" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1535" />
-			</output>
-		</layer>
-		<layer id="2077" name="__module.model.gpt_neox.layers.8.attention/aten::add_/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" names="1534,1538,1539,seq_len.33" />
-			</output>
-		</layer>
-		<layer id="2078" name="__module.model.gpt_neox.layers.8.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2079" name="__module.model.gpt_neox.layers.8.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2080" name="__module.model.gpt_neox.layers.8.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2081" name="__module.model.gpt_neox.layers.8.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1542,1545,cos.33">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2082" name="Constant_38245" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2083" name="__module.model.gpt_neox.layers.8.attention/aten::index/Gather" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="1548">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2084" name="__module.model.gpt_neox.layers.8.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1549,cos.35">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2085" name="__module.model.gpt_neox.layers.8.attention/aten::mul/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1553">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2086" name="Constant_272064" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2087" name="Constant_272063" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2088" name="__module.model.gpt_neox.layers.8.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2089" name="__module.model.gpt_neox.layers.8.attention/aten::size/Constant_5" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2090" name="__module.model.gpt_neox.layers.8.attention/aten::size/Gather_5" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1554,1559" />
-			</output>
-		</layer>
-		<layer id="2091" name="__module.model.gpt_neox.layers.8.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" />
-			</output>
-		</layer>
-		<layer id="2092" name="__module.model.gpt_neox.layers.8.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-			</input>
-			<output>
-				<port id="1" precision="I64" names="1556,1557,1561,1562" />
-			</output>
-		</layer>
-		<layer id="2093" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Constant_1" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2094" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2095" name="Constant_272062" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2096" name="ScatterUpdate_272065" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2097" name="Constant_272068" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2098" name="Constant_272071" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2099" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1563,x2.33">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2100" name="Constant_299156_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2101" name="Constant_299156" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2102" name="__module.model.gpt_neox.layers.8.attention/aten::neg/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1564">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2103" name="Constant_272116" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2104" name="Constant_272121" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2105" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1558">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2106" name="__module.model.gpt_neox.layers.8.attention/aten::cat/Concat" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1566">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2107" name="self.model.gpt_neox.layers.8.attention.rotary_emb.sin_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201162356" size="81920" />
-			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.8.attention.rotary_emb.sin_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2108" name="self.model.gpt_neox.layers.8.attention.rotary_emb.sin_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2109" name="__module.model.gpt_neox.layers.8.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2110" name="__module.model.gpt_neox.layers.8.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2111" name="__module.model.gpt_neox.layers.8.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1543,1546,sin.33">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2112" name="Constant_38248" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2113" name="__module.model.gpt_neox.layers.8.attention/aten::index/Gather_1" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="1551">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2114" name="__module.model.gpt_neox.layers.8.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1552,sin.35">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2115" name="__module.model.gpt_neox.layers.8.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1567">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2116" name="__module.model.gpt_neox.layers.8.attention/aten::add/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1568">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2117" name="Constant_272167" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2118" name="Constant_272170" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2119" name="Constant_272173" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2120" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1530">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2121" name="__module.model.gpt_neox.layers.8.attention/aten::cat/Concat_1" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1586,1596,query.51,query.53">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2122" name="Constant_272179" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2123" name="Constant_272182" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2124" name="Constant_272185" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2125" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1531,k.17">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2126" name="__module.model.gpt_neox.layers.8.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1569">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2127" name="Constant_272189" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2128" name="Constant_272188" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2129" name="__module.model.gpt_neox.layers.8.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2130" name="__module.model.gpt_neox.layers.8.attention/aten::size/Constant_7" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2131" name="__module.model.gpt_neox.layers.8.attention/aten::size/Gather_7" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1570,1575" />
-			</output>
-		</layer>
-		<layer id="2132" name="__module.model.gpt_neox.layers.8.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" />
-			</output>
-		</layer>
-		<layer id="2133" name="__module.model.gpt_neox.layers.8.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-			</input>
-			<output>
-				<port id="1" precision="I64" names="1572,1573,1577,1578" />
-			</output>
-		</layer>
-		<layer id="2134" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2135" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2136" name="Constant_272187" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2137" name="ScatterUpdate_272190" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2138" name="Constant_272193" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2139" name="Constant_272196" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2140" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1579,x2.35">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2141" name="Constant_299157_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2142" name="Constant_299157" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2143" name="__module.model.gpt_neox.layers.8.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1580">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2144" name="Constant_272241" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2145" name="Constant_272246" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2146" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1574">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2147" name="__module.model.gpt_neox.layers.8.attention/aten::cat/Concat_2" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1582">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2148" name="__module.model.gpt_neox.layers.8.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1583">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2149" name="__module.model.gpt_neox.layers.8.attention/aten::add/Add_1" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1584">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2150" name="Constant_272291" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2151" name="Constant_272294" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2152" name="Constant_272297" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2153" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1532">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2154" name="__module.model.gpt_neox.layers.8.attention/aten::cat/Concat_3" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1588">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2155" name="__module.model.gpt_neox.layers.8.attention/aten::to/Convert_1" type="Concat" version="opset1">
-			<data axis="-2" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="present.8.key">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2156" name="Constant_337650_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="2157" name="Constant_337650" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="2158" name="Constant_337643" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2159" name="Constant_337645" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2160" name="Constant_337647" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2161" name="Concat_337648" type="Concat" version="opset1">
-			<data axis="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2162" name="Broadcast_337651" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2163" name="ReadValue_330547" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.8.valuepresent.8.value" variable_type="f32" variable_shape="?,32,?,80" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32" names="past_key_values.8.value">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2164" name="Constant_323436" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64" />
-			</output>
-		</layer>
-		<layer id="2165" name="Gather_323437" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I64" />
-			</input>
-			<output>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2166" name="Constant_272303" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2167" name="Constant_272306" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2168" name="Constant_272309" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2169" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1526">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2170" name="Constant_4899" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="1527">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2171" name="__module.model.gpt_neox.layers.8.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1528">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2172" name="__module.model.gpt_neox.layers.8.attention/aten::cat/Concat_5" type="Concat" version="opset1">
-			<data axis="-2" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="present.8.value">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2173" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Slice_11" type="Const" version="opset1">
-			<data element_type="boolean" shape="1, 1, 2048, 2048" offset="201244276" size="4194304" />
-			<output>
-				<port id="0" precision="BOOL" names="1605">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2174" name="Constant_272313" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2175" name="Constant_272312" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162298" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2176" name="__module.model.gpt_neox.layers.8.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2177" name="__module.model.gpt_neox.layers.8.attention/aten::size/Constant_8" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2178" name="__module.model.gpt_neox.layers.8.attention/aten::size/Gather_8" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1600" />
-			</output>
-		</layer>
-		<layer id="2179" name="__module.model.gpt_neox.layers.8.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2180" name="__module.model.gpt_neox.layers.8.attention/aten::size/Constant_9" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2181" name="__module.model.gpt_neox.layers.8.attention/aten::size/Gather_9" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1598" />
-			</output>
-		</layer>
-		<layer id="2182" name="__module.model.gpt_neox.layers.8.attention/aten::sub/Subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" names="1602,1603" />
-			</output>
-		</layer>
-		<layer id="2183" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Constant_4" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2184" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2185" name="Constant_272311" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2186" name="ScatterUpdate_272314" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2187" name="Constant_272315" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2188" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2189" name="ScatterUpdate_272316" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2190" name="Constant_272319" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438612" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2191" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>3</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="BOOL" names="1606">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2192" name="Constant_272392" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2193" name="Constant_272393" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2194" name="Constant_272389" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2195" name="Constant_272388" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2196" name="ScatterUpdate_272394" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2197" name="Constant_272397" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2198" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="BOOL" names="1607,1608">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2199" name="__module.model.gpt_neox.layers.8.attention/aten::where/Convert_1_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="2200" name="__module.model.gpt_neox.layers.8.attention/aten::where/Convert_1" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="2201" name="__module.model.gpt_neox.layers.8.attention/aten::full/Convert" type="Const" version="opset1">
-			<data element_type="f32" shape="" offset="205438636" size="4" />
-			<output>
-				<port id="0" precision="FP32" names="1595" />
-			</output>
-		</layer>
-		<layer id="2202" name="__module.model.gpt_neox.layers.8.attention/aten::where/Select" type="Select" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="1" precision="FP32" />
-				<port id="2" precision="FP32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="1609,causal_mask.33">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2203" name="Add_290386" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2204" name="ShapeOf_290387" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2205" name="Constant_254449" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2206" name="__module.model.gpt_neox.layers.8.attention/aten::size/Constant_10" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2207" name="__module.model.gpt_neox.layers.8.attention/aten::size/Gather_10" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1593">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2208" name="__module.model.gpt_neox.layers.8.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1">
-			<data axis="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64" names="1610">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2209" name="__module.model.gpt_neox.layers.8.attention/aten::expand/Abs" type="Abs" version="opset1">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2210" name="Maximum_290388" type="Maximum" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2211" name="__module.model.gpt_neox.layers.8.attention/aten::add/Add_2" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1612">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2212" name="__module.model.gpt_neox.layers.8.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
-			<data causal="false" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1613,1614,sdpa_result.17,tensor.33">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2213" name="Constant_5340" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="1615">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2214" name="__module.model.gpt_neox.layers.8.attention/aten::permute/Transpose_3" type="Transpose" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1616">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2215" name="__module.model.gpt_neox.layers.8.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438650" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2216" name="__module.model.gpt_neox.layers.8.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
-			<data special_zero="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>3</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1621">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2217" name="self.model.gpt_neox.layers.8.attention.dense.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 2560" offset="835669714" size="6553600" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2218" name="Convert_21754" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2219" name="self.model.gpt_neox.layers.8.attention.dense.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="842223314" size="2560" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2220" name="Convert_21757" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2221" name="self.model.gpt_neox.layers.8.attention.dense.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2222" name="self.model.gpt_neox.layers.8.attention.dense.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="842225874" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2223" name="self.model.gpt_neox.layers.8.attention.dense.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2224" name="self.model.gpt_neox.layers.8.attention.dense.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2225" name="__module.model.gpt_neox.layers.8.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2226" name="Constant_299158_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="842230994" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2227" name="Constant_299158" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2228" name="__module.model.gpt_neox.layers.8.attention.dense/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1624,1626,input.35">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2229" name="__module.model.gpt_neox.layers.8/aten::add/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1644">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2230" name="__module.model.gpt_neox.layers.8/aten::add/Add_1" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1645,1647,hidden_states.19">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2231" name="__module.model.gpt_neox.layers.9.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867670" size="4" />
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2232" name="__module.model.gpt_neox.layers.9.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
-			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2233" name="Constant_299163_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="842236114" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2234" name="Constant_299163" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2235" name="__module.model.gpt_neox.layers.9.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2236" name="Constant_299164_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="842241234" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2237" name="Constant_299164" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2238" name="__module.model.gpt_neox.layers.9.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1657">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2239" name="self.model.gpt_neox.layers.9.attention.query_key_value.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 2560" offset="842246354" size="19660800" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2240" name="Convert_21784" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2241" name="self.model.gpt_neox.layers.9.attention.query_key_value.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 1" offset="861907154" size="7680" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2242" name="Convert_21787" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2243" name="self.model.gpt_neox.layers.9.attention.query_key_value.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2244" name="self.model.gpt_neox.layers.9.attention.query_key_value.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="7680, 1" offset="861914834" size="15360" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2245" name="self.model.gpt_neox.layers.9.attention.query_key_value.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2246" name="self.model.gpt_neox.layers.9.attention.query_key_value.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2247" name="__module.model.gpt_neox.layers.9.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2248" name="Constant_299165_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 7680" offset="861930194" size="15360" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2249" name="Constant_299165" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2250" name="__module.model.gpt_neox.layers.9.attention.query_key_value/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1664,qkv.37">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2251" name="__module.model.gpt_neox.layers.9.attention/prim::ListConstruct/Concat" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080154" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2252" name="__module.model.gpt_neox.layers.9.attention/aten::view/Reshape" type="Reshape" version="opset1">
-			<data special_zero="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1668,qkv.39">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2253" name="Constant_272678" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2254" name="Constant_272681" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2255" name="Constant_272684" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2256" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1675">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2257" name="Constant_5480" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="1676">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2258" name="__module.model.gpt_neox.layers.9.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1677">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2259" name="__module.model.gpt_neox.layers.9.attention/aten::cat/Concat_5" type="Concat" version="opset1">
-			<data axis="-2" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="present.9.value">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2261" name="Constant_337663_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="2262" name="Constant_337663" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="2263" name="Constant_337656" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2264" name="Constant_337658" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2265" name="Constant_337660" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2266" name="Concat_337661" type="Concat" version="opset1">
-			<data axis="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2267" name="Broadcast_337664" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2268" name="ReadValue_330549" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.9.keypresent.9.key" variable_type="f32" variable_shape="?,32,?,80" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32" names="past_key_values.9.key">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2269" name="Constant_323439" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64" />
-			</output>
-		</layer>
-		<layer id="2270" name="Gather_323440" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I64" />
-			</input>
-			<output>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2271" name="Constant_272428" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2272" name="Constant_272431" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2273" name="Constant_272434" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2274" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1672">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2275" name="Constant_5466" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="1673">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2276" name="__module.model.gpt_neox.layers.9.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1674,key.55">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2277" name="Constant_272554" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2278" name="Constant_272557" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2279" name="Constant_272560" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2280" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1680,k.19">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2281" name="self.model.gpt_neox.layers.9.attention.rotary_emb.cos_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201080346" size="81920" />
-			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.9.attention.rotary_emb.cos_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2282" name="self.model.gpt_neox.layers.9.attention.rotary_emb.cos_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2283" name="__module.model.gpt_neox.layers.9.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2284" name="__module.model.gpt_neox.layers.9.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2285" name="__module.model.gpt_neox.layers.9.attention/aten::size/Constant_2" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2286" name="__module.model.gpt_neox.layers.9.attention/aten::size/Gather_2" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1682" />
-			</output>
-		</layer>
-		<layer id="2287" name="__module.model.gpt_neox.layers.9.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2288" name="__module.model.gpt_neox.layers.9.attention/aten::size/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2289" name="__module.model.gpt_neox.layers.9.attention/aten::size/Gather_3" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1684" />
-			</output>
-		</layer>
-		<layer id="2290" name="__module.model.gpt_neox.layers.9.attention/aten::add_/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" names="1683,1687,1688,seq_len.37" />
-			</output>
-		</layer>
-		<layer id="2291" name="__module.model.gpt_neox.layers.9.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2292" name="__module.model.gpt_neox.layers.9.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2293" name="__module.model.gpt_neox.layers.9.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2294" name="__module.model.gpt_neox.layers.9.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1691,1694,cos.37">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2295" name="Constant_38251" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2296" name="__module.model.gpt_neox.layers.9.attention/aten::index/Gather" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="1697">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2297" name="__module.model.gpt_neox.layers.9.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1698,cos.39">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2298" name="__module.model.gpt_neox.layers.9.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1718">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2299" name="Constant_272564" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2300" name="Constant_272563" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2301" name="__module.model.gpt_neox.layers.9.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2302" name="__module.model.gpt_neox.layers.9.attention/aten::size/Constant_7" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2303" name="__module.model.gpt_neox.layers.9.attention/aten::size/Gather_7" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1719,1724" />
-			</output>
-		</layer>
-		<layer id="2304" name="__module.model.gpt_neox.layers.9.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" />
-			</output>
-		</layer>
-		<layer id="2305" name="__module.model.gpt_neox.layers.9.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-			</input>
-			<output>
-				<port id="1" precision="I64" names="1721,1722,1726,1727" />
-			</output>
-		</layer>
-		<layer id="2306" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2307" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2308" name="Constant_272562" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2309" name="ScatterUpdate_272565" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2310" name="Constant_272568" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2311" name="Constant_272571" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2312" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1728,x2.39">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2313" name="Constant_299167_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2314" name="Constant_299167" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2315" name="__module.model.gpt_neox.layers.9.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1729">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2316" name="Constant_272616" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2317" name="Constant_272621" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2318" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1723">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2319" name="__module.model.gpt_neox.layers.9.attention/aten::cat/Concat_2" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1731">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2320" name="self.model.gpt_neox.layers.9.attention.rotary_emb.sin_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201162356" size="81920" />
-			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.9.attention.rotary_emb.sin_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2321" name="self.model.gpt_neox.layers.9.attention.rotary_emb.sin_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2322" name="__module.model.gpt_neox.layers.9.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2323" name="__module.model.gpt_neox.layers.9.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2324" name="__module.model.gpt_neox.layers.9.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1692,1695,sin.37">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2325" name="Constant_38254" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2326" name="__module.model.gpt_neox.layers.9.attention/aten::index/Gather_1" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="1700">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2327" name="__module.model.gpt_neox.layers.9.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1701,sin.39">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2328" name="__module.model.gpt_neox.layers.9.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1732">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2329" name="__module.model.gpt_neox.layers.9.attention/aten::add/Add_1" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1733">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2330" name="Constant_272666" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2331" name="Constant_272669" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2332" name="Constant_272672" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2333" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1681">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2334" name="__module.model.gpt_neox.layers.9.attention/aten::cat/Concat_3" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1737">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2335" name="__module.model.gpt_neox.layers.9.attention/aten::to/Convert_1" type="Concat" version="opset1">
-			<data axis="-2" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="present.9.key">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2347" name="Constant_337676_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="2348" name="Constant_337676" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="2349" name="Constant_337669" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2350" name="Constant_337671" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2351" name="Constant_337673" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2352" name="Concat_337674" type="Concat" version="opset1">
-			<data axis="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2353" name="Broadcast_337677" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2354" name="ReadValue_330527" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.31.valuepresent.31.value" variable_type="f32" variable_shape="?,32,?,80" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32" names="past_key_values.31.value">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2355" name="Constant_323574" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64" />
-			</output>
-		</layer>
-		<layer id="2356" name="Gather_323575" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I64" />
-			</input>
-			<output>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2357" name="Constant_299159_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="861945554" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2358" name="Constant_299159" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2359" name="__module.model.gpt_neox.layers.9.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2360" name="Constant_299160_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="861950674" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2361" name="Constant_299160" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2362" name="__module.model.gpt_neox.layers.9.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1782">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2363" name="self.model.gpt_neox.layers.9.mlp.dense_h_to_4h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 2560" offset="861955794" size="26214400" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2364" name="Convert_21804" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2365" name="self.model.gpt_neox.layers.9.mlp.dense_h_to_4h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 1" offset="888170194" size="10240" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2366" name="Convert_21807" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2367" name="self.model.gpt_neox.layers.9.mlp.dense_h_to_4h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2368" name="self.model.gpt_neox.layers.9.mlp.dense_h_to_4h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="10240, 1" offset="888180434" size="20480" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2369" name="self.model.gpt_neox.layers.9.mlp.dense_h_to_4h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2370" name="self.model.gpt_neox.layers.9.mlp.dense_h_to_4h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2371" name="__module.model.gpt_neox.layers.9.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2372" name="Constant_299161_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 10240" offset="888200914" size="20480" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2373" name="Constant_299161" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2374" name="__module.model.gpt_neox.layers.9.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1787">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2375" name="__module.model.gpt_neox.layers.9.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
-			<data approximation_mode="ERF" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32" names="1788">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2376" name="self.model.gpt_neox.layers.9.mlp.dense_4h_to_h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 10240" offset="888221394" size="26214400" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2377" name="Convert_21814" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2378" name="self.model.gpt_neox.layers.9.mlp.dense_4h_to_h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="914435794" size="2560" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2379" name="Convert_21817" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2380" name="self.model.gpt_neox.layers.9.mlp.dense_4h_to_h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2381" name="self.model.gpt_neox.layers.9.mlp.dense_4h_to_h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="914438354" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2382" name="self.model.gpt_neox.layers.9.mlp.dense_4h_to_h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2383" name="self.model.gpt_neox.layers.9.mlp.dense_4h_to_h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2384" name="__module.model.gpt_neox.layers.9.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2385" name="Constant_299162_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="914443474" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2386" name="Constant_299162" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2387" name="__module.model.gpt_neox.layers.9.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1791,input.41">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2388" name="Constant_272404" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2389" name="Constant_272407" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2390" name="Constant_272410" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2391" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1669">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2392" name="Constant_5452" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="1670">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2393" name="__module.model.gpt_neox.layers.9.attention/aten::permute/Transpose" type="Transpose" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1671,query.55">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2394" name="Constant_272416" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2395" name="Constant_272419" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2396" name="Constant_272422" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2397" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1678,q.19">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2398" name="__module.model.gpt_neox.layers.9.attention/aten::mul/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1702">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2399" name="Constant_272439" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2400" name="Constant_272438" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2401" name="__module.model.gpt_neox.layers.9.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2402" name="__module.model.gpt_neox.layers.9.attention/aten::size/Constant_5" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2403" name="__module.model.gpt_neox.layers.9.attention/aten::size/Gather_5" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1703,1708" />
-			</output>
-		</layer>
-		<layer id="2404" name="__module.model.gpt_neox.layers.9.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" />
-			</output>
-		</layer>
-		<layer id="2405" name="__module.model.gpt_neox.layers.9.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-			</input>
-			<output>
-				<port id="1" precision="I64" names="1705,1706,1710,1711" />
-			</output>
-		</layer>
-		<layer id="2406" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Constant_1" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2407" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2408" name="Constant_272437" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2409" name="ScatterUpdate_272440" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2410" name="Constant_272443" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2411" name="Constant_272446" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2412" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1712,x2.37">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2413" name="Constant_299166_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2414" name="Constant_299166" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2415" name="__module.model.gpt_neox.layers.9.attention/aten::neg/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1713">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2416" name="Constant_272491" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2417" name="Constant_272496" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2418" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1707">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2419" name="__module.model.gpt_neox.layers.9.attention/aten::cat/Concat" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1715">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2420" name="__module.model.gpt_neox.layers.9.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1716">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2421" name="__module.model.gpt_neox.layers.9.attention/aten::add/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1717">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2422" name="Constant_272542" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2423" name="Constant_272545" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2424" name="Constant_272548" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2425" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1679">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2426" name="__module.model.gpt_neox.layers.9.attention/aten::cat/Concat_1" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1735,1745,query.57,query.59">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2427" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Slice_11" type="Const" version="opset1">
-			<data element_type="boolean" shape="1, 1, 2048, 2048" offset="201244276" size="4194304" />
-			<output>
-				<port id="0" precision="BOOL" names="1754">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2428" name="Constant_272688" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2429" name="Constant_272687" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162298" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2430" name="__module.model.gpt_neox.layers.9.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2431" name="__module.model.gpt_neox.layers.9.attention/aten::size/Constant_8" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2432" name="__module.model.gpt_neox.layers.9.attention/aten::size/Gather_8" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1749" />
-			</output>
-		</layer>
-		<layer id="2433" name="__module.model.gpt_neox.layers.9.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2434" name="__module.model.gpt_neox.layers.9.attention/aten::size/Constant_9" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2435" name="__module.model.gpt_neox.layers.9.attention/aten::size/Gather_9" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1747" />
-			</output>
-		</layer>
-		<layer id="2436" name="__module.model.gpt_neox.layers.9.attention/aten::sub/Subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" names="1751,1752" />
-			</output>
-		</layer>
-		<layer id="2437" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Constant_4" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2438" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2439" name="Constant_272686" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2440" name="ScatterUpdate_272689" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2441" name="Constant_272690" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2442" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2443" name="ScatterUpdate_272691" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2444" name="Constant_272694" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438612" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2445" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>3</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="BOOL" names="1755">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2446" name="Constant_272767" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2447" name="Constant_272768" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2448" name="Constant_272764" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2449" name="Constant_272763" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2450" name="ScatterUpdate_272769" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2451" name="Constant_272772" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2452" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="BOOL" names="1756,1757">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2453" name="__module.model.gpt_neox.layers.9.attention/aten::where/Convert_1_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="2454" name="__module.model.gpt_neox.layers.9.attention/aten::where/Convert_1" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="2455" name="__module.model.gpt_neox.layers.9.attention/aten::full/Convert" type="Const" version="opset1">
-			<data element_type="f32" shape="" offset="205438636" size="4" />
-			<output>
-				<port id="0" precision="FP32" names="1744" />
-			</output>
-		</layer>
-		<layer id="2456" name="__module.model.gpt_neox.layers.9.attention/aten::where/Select" type="Select" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="1" precision="FP32" />
-				<port id="2" precision="FP32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="1758,causal_mask.37">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2457" name="Add_290392" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2458" name="ShapeOf_290393" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2459" name="Constant_254464" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2460" name="__module.model.gpt_neox.layers.9.attention/aten::size/Constant_10" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2461" name="__module.model.gpt_neox.layers.9.attention/aten::size/Gather_10" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1742">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2462" name="__module.model.gpt_neox.layers.9.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1">
-			<data axis="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64" names="1759">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2463" name="__module.model.gpt_neox.layers.9.attention/aten::expand/Abs" type="Abs" version="opset1">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2464" name="Maximum_290394" type="Maximum" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2465" name="__module.model.gpt_neox.layers.9.attention/aten::add/Add_2" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1761">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2466" name="__module.model.gpt_neox.layers.9.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
-			<data causal="false" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1762,1763,sdpa_result.19,tensor.37">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2467" name="Constant_5921" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="1764">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2468" name="__module.model.gpt_neox.layers.9.attention/aten::permute/Transpose_3" type="Transpose" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1765">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2469" name="__module.model.gpt_neox.layers.9.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438650" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2470" name="__module.model.gpt_neox.layers.9.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
-			<data special_zero="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>3</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1770">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2471" name="self.model.gpt_neox.layers.9.attention.dense.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 2560" offset="914448594" size="6553600" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2472" name="Convert_21794" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2473" name="self.model.gpt_neox.layers.9.attention.dense.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="921002194" size="2560" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2474" name="Convert_21797" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2475" name="self.model.gpt_neox.layers.9.attention.dense.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2476" name="self.model.gpt_neox.layers.9.attention.dense.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="921004754" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2477" name="self.model.gpt_neox.layers.9.attention.dense.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2478" name="self.model.gpt_neox.layers.9.attention.dense.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2479" name="__module.model.gpt_neox.layers.9.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2480" name="Constant_299168_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="921009874" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2481" name="Constant_299168" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2482" name="__module.model.gpt_neox.layers.9.attention.dense/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1773,1775,input.39">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2483" name="__module.model.gpt_neox.layers.9/aten::add/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1793">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2484" name="__module.model.gpt_neox.layers.9/aten::add/Add_1" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1794,1796,hidden_states.21">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2485" name="__module.model.gpt_neox.layers.10.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867670" size="4" />
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2486" name="__module.model.gpt_neox.layers.10.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
-			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2487" name="Constant_299169_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="921014994" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2488" name="Constant_299169" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2489" name="__module.model.gpt_neox.layers.10.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2490" name="Constant_299170_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="921020114" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2491" name="Constant_299170" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2492" name="__module.model.gpt_neox.layers.10.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1931">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2493" name="self.model.gpt_neox.layers.10.mlp.dense_h_to_4h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 2560" offset="921025234" size="26214400" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2494" name="Convert_21844" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2495" name="self.model.gpt_neox.layers.10.mlp.dense_h_to_4h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 1" offset="947239634" size="10240" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2496" name="Convert_21847" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2497" name="self.model.gpt_neox.layers.10.mlp.dense_h_to_4h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2498" name="self.model.gpt_neox.layers.10.mlp.dense_h_to_4h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="10240, 1" offset="947249874" size="20480" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2499" name="self.model.gpt_neox.layers.10.mlp.dense_h_to_4h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2500" name="self.model.gpt_neox.layers.10.mlp.dense_h_to_4h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2501" name="__module.model.gpt_neox.layers.10.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2502" name="Constant_299171_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 10240" offset="947270354" size="20480" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2503" name="Constant_299171" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2504" name="__module.model.gpt_neox.layers.10.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1936">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2505" name="__module.model.gpt_neox.layers.10.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
-			<data approximation_mode="ERF" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32" names="1937">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2506" name="self.model.gpt_neox.layers.10.mlp.dense_4h_to_h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 10240" offset="947290834" size="26214400" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2507" name="Convert_21854" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2508" name="self.model.gpt_neox.layers.10.mlp.dense_4h_to_h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="973505234" size="2560" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2509" name="Convert_21857" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2510" name="self.model.gpt_neox.layers.10.mlp.dense_4h_to_h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2511" name="self.model.gpt_neox.layers.10.mlp.dense_4h_to_h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="973507794" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2512" name="self.model.gpt_neox.layers.10.mlp.dense_4h_to_h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2513" name="self.model.gpt_neox.layers.10.mlp.dense_4h_to_h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2514" name="__module.model.gpt_neox.layers.10.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2515" name="Constant_299172_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="973512914" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2516" name="Constant_299172" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2517" name="__module.model.gpt_neox.layers.10.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1940,input.45">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2518" name="Constant_299173_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="973518034" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2519" name="Constant_299173" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2520" name="__module.model.gpt_neox.layers.10.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2521" name="Constant_299174_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="973523154" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2522" name="Constant_299174" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2523" name="__module.model.gpt_neox.layers.10.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1806">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2524" name="self.model.gpt_neox.layers.10.attention.query_key_value.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 2560" offset="973528274" size="19660800" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2525" name="Convert_21824" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2526" name="self.model.gpt_neox.layers.10.attention.query_key_value.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 1" offset="993189074" size="7680" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2527" name="Convert_21827" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2528" name="self.model.gpt_neox.layers.10.attention.query_key_value.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2529" name="self.model.gpt_neox.layers.10.attention.query_key_value.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="7680, 1" offset="993196754" size="15360" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2530" name="self.model.gpt_neox.layers.10.attention.query_key_value.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2531" name="self.model.gpt_neox.layers.10.attention.query_key_value.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2532" name="__module.model.gpt_neox.layers.10.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2533" name="Constant_299175_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 7680" offset="993212114" size="15360" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2534" name="Constant_299175" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2535" name="__module.model.gpt_neox.layers.10.attention.query_key_value/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1813,qkv.41">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2536" name="__module.model.gpt_neox.layers.10.attention/prim::ListConstruct/Concat" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080154" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2537" name="__module.model.gpt_neox.layers.10.attention/aten::view/Reshape" type="Reshape" version="opset1">
-			<data special_zero="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1817,qkv.43">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2538" name="Constant_272779" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2539" name="Constant_272782" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2540" name="Constant_272785" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2541" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1818">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2542" name="Constant_6033" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="1819">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2543" name="__module.model.gpt_neox.layers.10.attention/aten::permute/Transpose" type="Transpose" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1820,query.61">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2544" name="Constant_272791" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2545" name="Constant_272794" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2546" name="Constant_272797" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2547" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1827,q.21">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2548" name="self.model.gpt_neox.layers.10.attention.rotary_emb.cos_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201080346" size="81920" />
-			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.10.attention.rotary_emb.cos_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2549" name="self.model.gpt_neox.layers.10.attention.rotary_emb.cos_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2550" name="__module.model.gpt_neox.layers.10.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2551" name="Constant_272803" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2552" name="Constant_272806" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2553" name="Constant_272809" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2554" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1821">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2555" name="Constant_6047" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="1822">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2556" name="__module.model.gpt_neox.layers.10.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1823,key.61">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2557" name="__module.model.gpt_neox.layers.10.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2558" name="__module.model.gpt_neox.layers.10.attention/aten::size/Constant_2" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2559" name="__module.model.gpt_neox.layers.10.attention/aten::size/Gather_2" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1831" />
-			</output>
-		</layer>
-		<layer id="2560" name="Constant_337689_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="2561" name="Constant_337689" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="2562" name="Constant_337682" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2563" name="Constant_337684" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2564" name="Constant_337686" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2565" name="Concat_337687" type="Concat" version="opset1">
-			<data axis="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2566" name="Broadcast_337690" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2567" name="ReadValue_330433" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.10.keypresent.10.key" variable_type="f32" variable_shape="?,32,?,80" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32" names="past_key_values.10.key">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2568" name="Constant_323445" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64" />
-			</output>
-		</layer>
-		<layer id="2569" name="Gather_323446" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I64" />
-			</input>
-			<output>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2570" name="__module.model.gpt_neox.layers.10.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2571" name="__module.model.gpt_neox.layers.10.attention/aten::size/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2572" name="__module.model.gpt_neox.layers.10.attention/aten::size/Gather_3" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1833" />
-			</output>
-		</layer>
-		<layer id="2573" name="__module.model.gpt_neox.layers.10.attention/aten::add_/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" names="1832,1836,1837,seq_len.41" />
-			</output>
-		</layer>
-		<layer id="2574" name="__module.model.gpt_neox.layers.10.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2575" name="__module.model.gpt_neox.layers.10.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2576" name="__module.model.gpt_neox.layers.10.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2577" name="__module.model.gpt_neox.layers.10.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1840,1843,cos.41">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2578" name="Constant_38257" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2579" name="__module.model.gpt_neox.layers.10.attention/aten::index/Gather" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="1846">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2580" name="__module.model.gpt_neox.layers.10.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1847,cos.43">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2581" name="__module.model.gpt_neox.layers.10.attention/aten::mul/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1851">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2582" name="Constant_272814" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2583" name="Constant_272813" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2584" name="__module.model.gpt_neox.layers.10.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2585" name="__module.model.gpt_neox.layers.10.attention/aten::size/Constant_5" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2586" name="__module.model.gpt_neox.layers.10.attention/aten::size/Gather_5" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1852,1857" />
-			</output>
-		</layer>
-		<layer id="2587" name="__module.model.gpt_neox.layers.10.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" />
-			</output>
-		</layer>
-		<layer id="2588" name="__module.model.gpt_neox.layers.10.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-			</input>
-			<output>
-				<port id="1" precision="I64" names="1854,1855,1859,1860" />
-			</output>
-		</layer>
-		<layer id="2589" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Constant_1" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2590" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2591" name="Constant_272812" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2592" name="ScatterUpdate_272815" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2593" name="Constant_272818" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2594" name="Constant_272821" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2595" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1861,x2.41">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2596" name="Constant_299176_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2597" name="Constant_299176" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2598" name="__module.model.gpt_neox.layers.10.attention/aten::neg/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1862">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2599" name="Constant_272866" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2600" name="Constant_272871" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2601" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1856">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2602" name="__module.model.gpt_neox.layers.10.attention/aten::cat/Concat" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1864">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2603" name="self.model.gpt_neox.layers.10.attention.rotary_emb.sin_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201162356" size="81920" />
-			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.10.attention.rotary_emb.sin_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2604" name="self.model.gpt_neox.layers.10.attention.rotary_emb.sin_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2605" name="__module.model.gpt_neox.layers.10.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2606" name="__module.model.gpt_neox.layers.10.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2607" name="__module.model.gpt_neox.layers.10.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1841,1844,sin.41">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2608" name="Constant_38260" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2609" name="__module.model.gpt_neox.layers.10.attention/aten::index/Gather_1" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="1849">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2610" name="__module.model.gpt_neox.layers.10.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1850,sin.43">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2611" name="__module.model.gpt_neox.layers.10.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1865">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2612" name="__module.model.gpt_neox.layers.10.attention/aten::add/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1866">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2613" name="Constant_272917" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2614" name="Constant_272920" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2615" name="Constant_272923" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2616" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1828">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2617" name="__module.model.gpt_neox.layers.10.attention/aten::cat/Concat_1" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1884,1894,query.63,query.65">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2618" name="Constant_272929" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2619" name="Constant_272932" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2620" name="Constant_272935" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2621" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1829,k.21">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2622" name="__module.model.gpt_neox.layers.10.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1867">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2623" name="Constant_272939" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2624" name="Constant_272938" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2625" name="__module.model.gpt_neox.layers.10.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2626" name="__module.model.gpt_neox.layers.10.attention/aten::size/Constant_7" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2627" name="__module.model.gpt_neox.layers.10.attention/aten::size/Gather_7" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1868,1873" />
-			</output>
-		</layer>
-		<layer id="2628" name="__module.model.gpt_neox.layers.10.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" />
-			</output>
-		</layer>
-		<layer id="2629" name="__module.model.gpt_neox.layers.10.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-			</input>
-			<output>
-				<port id="1" precision="I64" names="1870,1871,1875,1876" />
-			</output>
-		</layer>
-		<layer id="2630" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2631" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2632" name="Constant_272937" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2633" name="ScatterUpdate_272940" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2634" name="Constant_272943" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2635" name="Constant_272946" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2636" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1877,x2.43">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2637" name="Constant_299177_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2638" name="Constant_299177" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2639" name="__module.model.gpt_neox.layers.10.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1878">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2640" name="Constant_272991" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2641" name="Constant_272996" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2642" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1872">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2643" name="__module.model.gpt_neox.layers.10.attention/aten::cat/Concat_2" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1880">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2644" name="__module.model.gpt_neox.layers.10.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1881">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2645" name="__module.model.gpt_neox.layers.10.attention/aten::add/Add_1" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1882">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2646" name="Constant_273041" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2647" name="Constant_273044" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2648" name="Constant_273047" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2649" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1830">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2650" name="__module.model.gpt_neox.layers.10.attention/aten::cat/Concat_3" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1886">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2651" name="__module.model.gpt_neox.layers.10.attention/aten::to/Convert_1" type="Concat" version="opset1">
-			<data axis="-2" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="present.10.key">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2652" name="Constant_337702_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="2653" name="Constant_337702" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="2654" name="Constant_337695" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2655" name="Constant_337697" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2656" name="Constant_337699" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2657" name="Concat_337700" type="Concat" version="opset1">
-			<data axis="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2658" name="Broadcast_337703" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2659" name="ReadValue_330435" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.10.valuepresent.10.value" variable_type="f32" variable_shape="?,32,?,80" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32" names="past_key_values.10.value">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2660" name="Constant_323448" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64" />
-			</output>
-		</layer>
-		<layer id="2661" name="Gather_323449" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I64" />
-			</input>
-			<output>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2662" name="Constant_273053" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2663" name="Constant_273056" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2664" name="Constant_273059" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2665" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1824">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2666" name="Constant_6061" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="1825">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2667" name="__module.model.gpt_neox.layers.10.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1826">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2668" name="__module.model.gpt_neox.layers.10.attention/aten::cat/Concat_5" type="Concat" version="opset1">
-			<data axis="-2" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="present.10.value">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2669" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Slice_11" type="Const" version="opset1">
-			<data element_type="boolean" shape="1, 1, 2048, 2048" offset="201244276" size="4194304" />
-			<output>
-				<port id="0" precision="BOOL" names="1903">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2670" name="Constant_273063" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2671" name="Constant_273062" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162298" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2672" name="__module.model.gpt_neox.layers.10.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2673" name="__module.model.gpt_neox.layers.10.attention/aten::size/Constant_8" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2674" name="__module.model.gpt_neox.layers.10.attention/aten::size/Gather_8" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1898" />
-			</output>
-		</layer>
-		<layer id="2675" name="__module.model.gpt_neox.layers.10.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2676" name="__module.model.gpt_neox.layers.10.attention/aten::size/Constant_9" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2677" name="__module.model.gpt_neox.layers.10.attention/aten::size/Gather_9" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1896" />
-			</output>
-		</layer>
-		<layer id="2678" name="__module.model.gpt_neox.layers.10.attention/aten::sub/Subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" names="1900,1901" />
-			</output>
-		</layer>
-		<layer id="2679" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Constant_4" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2680" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2681" name="Constant_273061" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2682" name="ScatterUpdate_273064" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2683" name="Constant_273065" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2684" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2685" name="ScatterUpdate_273066" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2686" name="Constant_273069" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438612" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2687" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>3</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="BOOL" names="1904">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2688" name="Constant_273142" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2689" name="Constant_273143" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2690" name="Constant_273139" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2691" name="Constant_273138" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2692" name="ScatterUpdate_273144" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2693" name="Constant_273147" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2694" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="BOOL" names="1905,1906">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2695" name="__module.model.gpt_neox.layers.10.attention/aten::where/Convert_1_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="2696" name="__module.model.gpt_neox.layers.10.attention/aten::where/Convert_1" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="2697" name="__module.model.gpt_neox.layers.10.attention/aten::full/Convert" type="Const" version="opset1">
-			<data element_type="f32" shape="" offset="205438636" size="4" />
-			<output>
-				<port id="0" precision="FP32" names="1893" />
-			</output>
-		</layer>
-		<layer id="2698" name="__module.model.gpt_neox.layers.10.attention/aten::where/Select" type="Select" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="1" precision="FP32" />
-				<port id="2" precision="FP32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="1907,causal_mask.41">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2699" name="Add_290398" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2700" name="ShapeOf_290399" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2701" name="Constant_254479" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2702" name="__module.model.gpt_neox.layers.10.attention/aten::size/Constant_10" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2703" name="__module.model.gpt_neox.layers.10.attention/aten::size/Gather_10" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1891">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2704" name="__module.model.gpt_neox.layers.10.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1">
-			<data axis="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64" names="1908">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2705" name="__module.model.gpt_neox.layers.10.attention/aten::expand/Abs" type="Abs" version="opset1">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2706" name="Maximum_290400" type="Maximum" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2707" name="__module.model.gpt_neox.layers.10.attention/aten::add/Add_2" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1910">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2708" name="__module.model.gpt_neox.layers.10.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
-			<data causal="false" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1911,1912,sdpa_result.21,tensor.41">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2709" name="Constant_6502" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="1913">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2710" name="__module.model.gpt_neox.layers.10.attention/aten::permute/Transpose_3" type="Transpose" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1914">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2711" name="__module.model.gpt_neox.layers.10.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438650" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2712" name="__module.model.gpt_neox.layers.10.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
-			<data special_zero="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>3</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1919">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2713" name="self.model.gpt_neox.layers.10.attention.dense.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 2560" offset="993227474" size="6553600" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2714" name="Convert_21834" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2715" name="self.model.gpt_neox.layers.10.attention.dense.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="999781074" size="2560" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2716" name="Convert_21837" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2717" name="self.model.gpt_neox.layers.10.attention.dense.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2718" name="self.model.gpt_neox.layers.10.attention.dense.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="999783634" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2719" name="self.model.gpt_neox.layers.10.attention.dense.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2720" name="self.model.gpt_neox.layers.10.attention.dense.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2721" name="__module.model.gpt_neox.layers.10.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2722" name="Constant_299178_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="999788754" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2723" name="Constant_299178" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2724" name="__module.model.gpt_neox.layers.10.attention.dense/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1922,1924,input.43">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2725" name="__module.model.gpt_neox.layers.10/aten::add/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1942">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2726" name="__module.model.gpt_neox.layers.10/aten::add/Add_1" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1943,1945,hidden_states.23">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2727" name="__module.model.gpt_neox.layers.11.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867670" size="4" />
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2728" name="__module.model.gpt_neox.layers.11.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
-			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2729" name="Constant_299179_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="999793874" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2730" name="Constant_299179" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2731" name="__module.model.gpt_neox.layers.11.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2732" name="Constant_299180_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="999798994" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2733" name="Constant_299180" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2734" name="__module.model.gpt_neox.layers.11.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2080">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2735" name="self.model.gpt_neox.layers.11.mlp.dense_h_to_4h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 2560" offset="999804114" size="26214400" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2736" name="Convert_21884" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2737" name="self.model.gpt_neox.layers.11.mlp.dense_h_to_4h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 1" offset="1026018514" size="10240" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2738" name="Convert_21887" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2739" name="self.model.gpt_neox.layers.11.mlp.dense_h_to_4h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2740" name="self.model.gpt_neox.layers.11.mlp.dense_h_to_4h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="10240, 1" offset="1026028754" size="20480" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2741" name="self.model.gpt_neox.layers.11.mlp.dense_h_to_4h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2742" name="self.model.gpt_neox.layers.11.mlp.dense_h_to_4h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2743" name="__module.model.gpt_neox.layers.11.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2744" name="Constant_299181_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 10240" offset="1026049234" size="20480" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2745" name="Constant_299181" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2746" name="__module.model.gpt_neox.layers.11.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2085">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2747" name="__module.model.gpt_neox.layers.11.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
-			<data approximation_mode="ERF" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32" names="2086">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2748" name="self.model.gpt_neox.layers.11.mlp.dense_4h_to_h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 10240" offset="1026069714" size="26214400" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2749" name="Convert_21894" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2750" name="self.model.gpt_neox.layers.11.mlp.dense_4h_to_h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="1052284114" size="2560" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2751" name="Convert_21897" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2752" name="self.model.gpt_neox.layers.11.mlp.dense_4h_to_h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2753" name="self.model.gpt_neox.layers.11.mlp.dense_4h_to_h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="1052286674" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2754" name="self.model.gpt_neox.layers.11.mlp.dense_4h_to_h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2755" name="self.model.gpt_neox.layers.11.mlp.dense_4h_to_h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2756" name="__module.model.gpt_neox.layers.11.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2757" name="Constant_299182_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1052291794" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2758" name="Constant_299182" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2759" name="__module.model.gpt_neox.layers.11.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2089,input.49">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2760" name="Constant_299183_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1052296914" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2761" name="Constant_299183" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2762" name="__module.model.gpt_neox.layers.11.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2763" name="Constant_299184_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1052302034" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2764" name="Constant_299184" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2765" name="__module.model.gpt_neox.layers.11.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1955">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2766" name="self.model.gpt_neox.layers.11.attention.query_key_value.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 2560" offset="1052307154" size="19660800" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2767" name="Convert_21864" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2768" name="self.model.gpt_neox.layers.11.attention.query_key_value.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 1" offset="1071967954" size="7680" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2769" name="Convert_21867" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2770" name="self.model.gpt_neox.layers.11.attention.query_key_value.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2771" name="self.model.gpt_neox.layers.11.attention.query_key_value.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="7680, 1" offset="1071975634" size="15360" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2772" name="self.model.gpt_neox.layers.11.attention.query_key_value.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2773" name="self.model.gpt_neox.layers.11.attention.query_key_value.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2774" name="__module.model.gpt_neox.layers.11.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2775" name="Constant_299185_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 7680" offset="1071990994" size="15360" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2776" name="Constant_299185" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2777" name="__module.model.gpt_neox.layers.11.attention.query_key_value/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1962,qkv.45">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2778" name="__module.model.gpt_neox.layers.11.attention/prim::ListConstruct/Concat" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080154" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2779" name="__module.model.gpt_neox.layers.11.attention/aten::view/Reshape" type="Reshape" version="opset1">
-			<data special_zero="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1966,qkv.47">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2780" name="Constant_273154" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2781" name="Constant_273157" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2782" name="Constant_273160" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2783" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1967">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2784" name="Constant_6614" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="1968">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2785" name="__module.model.gpt_neox.layers.11.attention/aten::permute/Transpose" type="Transpose" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1969,query.67">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2786" name="Constant_273166" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2787" name="Constant_273169" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2788" name="Constant_273172" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2789" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1976,q.23">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2790" name="self.model.gpt_neox.layers.11.attention.rotary_emb.cos_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201080346" size="81920" />
-			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.11.attention.rotary_emb.cos_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2791" name="self.model.gpt_neox.layers.11.attention.rotary_emb.cos_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2792" name="__module.model.gpt_neox.layers.11.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2793" name="Constant_273178" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2794" name="Constant_273181" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2795" name="Constant_273184" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2796" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1970">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2797" name="Constant_6628" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="1971">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2798" name="__module.model.gpt_neox.layers.11.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1972,key.67">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2799" name="__module.model.gpt_neox.layers.11.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2800" name="__module.model.gpt_neox.layers.11.attention/aten::size/Constant_2" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2801" name="__module.model.gpt_neox.layers.11.attention/aten::size/Gather_2" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1980" />
-			</output>
-		</layer>
-		<layer id="2802" name="Constant_337715_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="2803" name="Constant_337715" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="2804" name="Constant_337708" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2805" name="Constant_337710" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2806" name="Constant_337712" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2807" name="Concat_337713" type="Concat" version="opset1">
-			<data axis="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2808" name="Broadcast_337716" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2809" name="ReadValue_330437" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.11.keypresent.11.key" variable_type="f32" variable_shape="?,32,?,80" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32" names="past_key_values.11.key">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2810" name="Constant_323451" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64" />
-			</output>
-		</layer>
-		<layer id="2811" name="Gather_323452" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I64" />
-			</input>
-			<output>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2812" name="__module.model.gpt_neox.layers.11.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2813" name="__module.model.gpt_neox.layers.11.attention/aten::size/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2814" name="__module.model.gpt_neox.layers.11.attention/aten::size/Gather_3" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="1982" />
-			</output>
-		</layer>
-		<layer id="2815" name="__module.model.gpt_neox.layers.11.attention/aten::add_/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" names="1981,1985,1986,seq_len.45" />
-			</output>
-		</layer>
-		<layer id="2816" name="__module.model.gpt_neox.layers.11.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2817" name="__module.model.gpt_neox.layers.11.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2818" name="__module.model.gpt_neox.layers.11.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2819" name="__module.model.gpt_neox.layers.11.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1989,1992,cos.45">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2820" name="Constant_38263" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2821" name="__module.model.gpt_neox.layers.11.attention/aten::index/Gather" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="1995">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2822" name="__module.model.gpt_neox.layers.11.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1996,cos.47">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2823" name="__module.model.gpt_neox.layers.11.attention/aten::mul/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2000">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2824" name="Constant_273189" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2825" name="Constant_273188" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2826" name="__module.model.gpt_neox.layers.11.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2827" name="__module.model.gpt_neox.layers.11.attention/aten::size/Constant_5" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2828" name="__module.model.gpt_neox.layers.11.attention/aten::size/Gather_5" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="2001,2006" />
-			</output>
-		</layer>
-		<layer id="2829" name="__module.model.gpt_neox.layers.11.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" />
-			</output>
-		</layer>
-		<layer id="2830" name="__module.model.gpt_neox.layers.11.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-			</input>
-			<output>
-				<port id="1" precision="I64" names="2003,2004,2008,2009" />
-			</output>
-		</layer>
-		<layer id="2831" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Constant_1" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2832" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2833" name="Constant_273187" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2834" name="ScatterUpdate_273190" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2835" name="Constant_273193" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2836" name="Constant_273196" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2837" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="2010,x2.45">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2838" name="Constant_299186_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2839" name="Constant_299186" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2840" name="__module.model.gpt_neox.layers.11.attention/aten::neg/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2011">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2841" name="Constant_273241" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2842" name="Constant_273246" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2843" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="2005">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2844" name="__module.model.gpt_neox.layers.11.attention/aten::cat/Concat" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2013">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2845" name="self.model.gpt_neox.layers.11.attention.rotary_emb.sin_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201162356" size="81920" />
-			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.11.attention.rotary_emb.sin_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2846" name="self.model.gpt_neox.layers.11.attention.rotary_emb.sin_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2847" name="__module.model.gpt_neox.layers.11.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2848" name="__module.model.gpt_neox.layers.11.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2849" name="__module.model.gpt_neox.layers.11.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1990,1993,sin.45">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2850" name="Constant_38266" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2851" name="__module.model.gpt_neox.layers.11.attention/aten::index/Gather_1" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="1998">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2852" name="__module.model.gpt_neox.layers.11.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1999,sin.47">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2853" name="__module.model.gpt_neox.layers.11.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2014">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2854" name="__module.model.gpt_neox.layers.11.attention/aten::add/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2015">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2855" name="Constant_273292" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2856" name="Constant_273295" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2857" name="Constant_273298" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2858" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1977">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2859" name="__module.model.gpt_neox.layers.11.attention/aten::cat/Concat_1" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2033,2043,query.69,query.71">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2860" name="Constant_273304" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2861" name="Constant_273307" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2862" name="Constant_273310" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2863" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1978,k.23">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2864" name="__module.model.gpt_neox.layers.11.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2016">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2865" name="Constant_273314" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2866" name="Constant_273313" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2867" name="__module.model.gpt_neox.layers.11.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2868" name="__module.model.gpt_neox.layers.11.attention/aten::size/Constant_7" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2869" name="__module.model.gpt_neox.layers.11.attention/aten::size/Gather_7" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="2017,2022" />
-			</output>
-		</layer>
-		<layer id="2870" name="__module.model.gpt_neox.layers.11.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" />
-			</output>
-		</layer>
-		<layer id="2871" name="__module.model.gpt_neox.layers.11.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-			</input>
-			<output>
-				<port id="1" precision="I64" names="2019,2020,2024,2025" />
-			</output>
-		</layer>
-		<layer id="2872" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2873" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2874" name="Constant_273312" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2875" name="ScatterUpdate_273315" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2876" name="Constant_273318" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2877" name="Constant_273321" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2878" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="2026,x2.47">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2879" name="Constant_299187_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2880" name="Constant_299187" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2881" name="__module.model.gpt_neox.layers.11.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2027">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2882" name="Constant_273366" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2883" name="Constant_273371" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2884" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="2021">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2885" name="__module.model.gpt_neox.layers.11.attention/aten::cat/Concat_2" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2029">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2886" name="__module.model.gpt_neox.layers.11.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2030">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2887" name="__module.model.gpt_neox.layers.11.attention/aten::add/Add_1" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2031">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2888" name="Constant_273416" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2889" name="Constant_273419" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2890" name="Constant_273422" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2891" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1979">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2892" name="__module.model.gpt_neox.layers.11.attention/aten::cat/Concat_3" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2035">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2893" name="__module.model.gpt_neox.layers.11.attention/aten::to/Convert_1" type="Concat" version="opset1">
-			<data axis="-2" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="present.11.key">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2894" name="Constant_337728_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="2895" name="Constant_337728" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="2896" name="Constant_337721" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2897" name="Constant_337723" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2898" name="Constant_337725" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2899" name="Concat_337726" type="Concat" version="opset1">
-			<data axis="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2900" name="Broadcast_337729" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2901" name="ReadValue_330439" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.11.valuepresent.11.value" variable_type="f32" variable_shape="?,32,?,80" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32" names="past_key_values.11.value">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2902" name="Constant_323454" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64" />
-			</output>
-		</layer>
-		<layer id="2903" name="Gather_323455" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I64" />
-			</input>
-			<output>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2904" name="Constant_273428" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2905" name="Constant_273431" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2906" name="Constant_273434" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2907" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="1973">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2908" name="Constant_6642" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="1974">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2909" name="__module.model.gpt_neox.layers.11.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="1975">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2910" name="__module.model.gpt_neox.layers.11.attention/aten::cat/Concat_5" type="Concat" version="opset1">
-			<data axis="-2" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="present.11.value">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2911" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Slice_11" type="Const" version="opset1">
-			<data element_type="boolean" shape="1, 1, 2048, 2048" offset="201244276" size="4194304" />
-			<output>
-				<port id="0" precision="BOOL" names="2052">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2912" name="Constant_273438" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2913" name="Constant_273437" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162298" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2914" name="__module.model.gpt_neox.layers.11.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2915" name="__module.model.gpt_neox.layers.11.attention/aten::size/Constant_8" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2916" name="__module.model.gpt_neox.layers.11.attention/aten::size/Gather_8" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="2047" />
-			</output>
-		</layer>
-		<layer id="2917" name="__module.model.gpt_neox.layers.11.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2918" name="__module.model.gpt_neox.layers.11.attention/aten::size/Constant_9" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2919" name="__module.model.gpt_neox.layers.11.attention/aten::size/Gather_9" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="2045" />
-			</output>
-		</layer>
-		<layer id="2920" name="__module.model.gpt_neox.layers.11.attention/aten::sub/Subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" names="2049,2050" />
-			</output>
-		</layer>
-		<layer id="2921" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Constant_4" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2922" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2923" name="Constant_273436" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2924" name="ScatterUpdate_273439" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2925" name="Constant_273440" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2926" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2927" name="ScatterUpdate_273441" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2928" name="Constant_273444" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438612" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2929" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>3</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="BOOL" names="2053">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2930" name="Constant_273517" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2931" name="Constant_273518" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2932" name="Constant_273514" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2933" name="Constant_273513" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2934" name="ScatterUpdate_273519" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2935" name="Constant_273522" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2936" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="BOOL" names="2054,2055">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2937" name="__module.model.gpt_neox.layers.11.attention/aten::where/Convert_1_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="2938" name="__module.model.gpt_neox.layers.11.attention/aten::where/Convert_1" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="2939" name="__module.model.gpt_neox.layers.11.attention/aten::full/Convert" type="Const" version="opset1">
-			<data element_type="f32" shape="" offset="205438636" size="4" />
-			<output>
-				<port id="0" precision="FP32" names="2042" />
-			</output>
-		</layer>
-		<layer id="2940" name="__module.model.gpt_neox.layers.11.attention/aten::where/Select" type="Select" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="1" precision="FP32" />
-				<port id="2" precision="FP32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="2056,causal_mask.45">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2941" name="Add_290404" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2942" name="ShapeOf_290405" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2943" name="Constant_254494" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2944" name="__module.model.gpt_neox.layers.11.attention/aten::size/Constant_10" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="2945" name="__module.model.gpt_neox.layers.11.attention/aten::size/Gather_10" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="2040">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2946" name="__module.model.gpt_neox.layers.11.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1">
-			<data axis="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64" names="2057">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2947" name="__module.model.gpt_neox.layers.11.attention/aten::expand/Abs" type="Abs" version="opset1">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2948" name="Maximum_290406" type="Maximum" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2949" name="__module.model.gpt_neox.layers.11.attention/aten::add/Add_2" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2059">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2950" name="__module.model.gpt_neox.layers.11.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
-			<data causal="false" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="2060,2061,sdpa_result.23,tensor.45">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2951" name="Constant_7083" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="2062">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2952" name="__module.model.gpt_neox.layers.11.attention/aten::permute/Transpose_3" type="Transpose" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2063">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2953" name="__module.model.gpt_neox.layers.11.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438650" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2954" name="__module.model.gpt_neox.layers.11.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
-			<data special_zero="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>3</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2068">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2955" name="self.model.gpt_neox.layers.11.attention.dense.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 2560" offset="1072006354" size="6553600" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2956" name="Convert_21874" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2957" name="self.model.gpt_neox.layers.11.attention.dense.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="1078559954" size="2560" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2958" name="Convert_21877" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2959" name="self.model.gpt_neox.layers.11.attention.dense.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2960" name="self.model.gpt_neox.layers.11.attention.dense.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="1078562514" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2961" name="self.model.gpt_neox.layers.11.attention.dense.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2962" name="self.model.gpt_neox.layers.11.attention.dense.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2963" name="__module.model.gpt_neox.layers.11.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2964" name="Constant_299188_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1078567634" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2965" name="Constant_299188" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2966" name="__module.model.gpt_neox.layers.11.attention.dense/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2071,2073,input.47">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2967" name="__module.model.gpt_neox.layers.11/aten::add/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2091">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2968" name="__module.model.gpt_neox.layers.11/aten::add/Add_1" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2092,2094,hidden_states.25">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2969" name="__module.model.gpt_neox.layers.12.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867670" size="4" />
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2970" name="__module.model.gpt_neox.layers.12.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
-			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2971" name="Constant_299189_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1078572754" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2972" name="Constant_299189" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2973" name="__module.model.gpt_neox.layers.12.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2974" name="Constant_299190_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1078577874" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2975" name="Constant_299190" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2976" name="__module.model.gpt_neox.layers.12.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2229">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2977" name="self.model.gpt_neox.layers.12.mlp.dense_h_to_4h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 2560" offset="1078582994" size="26214400" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2978" name="Convert_21924" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2979" name="self.model.gpt_neox.layers.12.mlp.dense_h_to_4h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 1" offset="1104797394" size="10240" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2980" name="Convert_21927" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2981" name="self.model.gpt_neox.layers.12.mlp.dense_h_to_4h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2982" name="self.model.gpt_neox.layers.12.mlp.dense_h_to_4h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="10240, 1" offset="1104807634" size="20480" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2983" name="self.model.gpt_neox.layers.12.mlp.dense_h_to_4h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2984" name="self.model.gpt_neox.layers.12.mlp.dense_h_to_4h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2985" name="__module.model.gpt_neox.layers.12.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2986" name="Constant_299191_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 10240" offset="1104828114" size="20480" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2987" name="Constant_299191" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2988" name="__module.model.gpt_neox.layers.12.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2234">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2989" name="__module.model.gpt_neox.layers.12.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
-			<data approximation_mode="ERF" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32" names="2235">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2990" name="self.model.gpt_neox.layers.12.mlp.dense_4h_to_h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 10240" offset="1104848594" size="26214400" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2991" name="Convert_21934" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2992" name="self.model.gpt_neox.layers.12.mlp.dense_4h_to_h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="1131062994" size="2560" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2993" name="Convert_21937" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2994" name="self.model.gpt_neox.layers.12.mlp.dense_4h_to_h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2995" name="self.model.gpt_neox.layers.12.mlp.dense_4h_to_h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="1131065554" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2996" name="self.model.gpt_neox.layers.12.mlp.dense_4h_to_h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2997" name="self.model.gpt_neox.layers.12.mlp.dense_4h_to_h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2998" name="__module.model.gpt_neox.layers.12.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="2999" name="Constant_299192_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1131070674" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3000" name="Constant_299192" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3001" name="__module.model.gpt_neox.layers.12.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2238,input.53">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3002" name="Constant_299193_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1131075794" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3003" name="Constant_299193" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3004" name="__module.model.gpt_neox.layers.12.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3005" name="Constant_299194_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1131080914" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3006" name="Constant_299194" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3007" name="__module.model.gpt_neox.layers.12.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2104">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3008" name="self.model.gpt_neox.layers.12.attention.query_key_value.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 2560" offset="1131086034" size="19660800" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3009" name="Convert_21904" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3010" name="self.model.gpt_neox.layers.12.attention.query_key_value.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 1" offset="1150746834" size="7680" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3011" name="Convert_21907" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3012" name="self.model.gpt_neox.layers.12.attention.query_key_value.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3013" name="self.model.gpt_neox.layers.12.attention.query_key_value.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="7680, 1" offset="1150754514" size="15360" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3014" name="self.model.gpt_neox.layers.12.attention.query_key_value.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3015" name="self.model.gpt_neox.layers.12.attention.query_key_value.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3016" name="__module.model.gpt_neox.layers.12.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3017" name="Constant_299195_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 7680" offset="1150769874" size="15360" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3018" name="Constant_299195" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3019" name="__module.model.gpt_neox.layers.12.attention.query_key_value/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2111,qkv.49">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3020" name="__module.model.gpt_neox.layers.12.attention/prim::ListConstruct/Concat" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080154" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3021" name="__module.model.gpt_neox.layers.12.attention/aten::view/Reshape" type="Reshape" version="opset1">
-			<data special_zero="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2115,qkv.51">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3022" name="Constant_273529" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3023" name="Constant_273532" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3024" name="Constant_273535" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3025" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="2116">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3026" name="Constant_7195" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="2117">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3027" name="__module.model.gpt_neox.layers.12.attention/aten::permute/Transpose" type="Transpose" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2118,query.73">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3028" name="Constant_273541" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3029" name="Constant_273544" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3030" name="Constant_273547" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3031" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="2125,q.25">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3032" name="self.model.gpt_neox.layers.12.attention.rotary_emb.cos_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201080346" size="81920" />
-			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.12.attention.rotary_emb.cos_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3033" name="self.model.gpt_neox.layers.12.attention.rotary_emb.cos_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3034" name="__module.model.gpt_neox.layers.12.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3035" name="Constant_273553" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3036" name="Constant_273556" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3037" name="Constant_273559" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3038" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="2119">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3039" name="Constant_7209" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="2120">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3040" name="__module.model.gpt_neox.layers.12.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2121,key.73">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3041" name="__module.model.gpt_neox.layers.12.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3042" name="__module.model.gpt_neox.layers.12.attention/aten::size/Constant_2" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="3043" name="__module.model.gpt_neox.layers.12.attention/aten::size/Gather_2" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="2129" />
-			</output>
-		</layer>
-		<layer id="3044" name="Constant_337741_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="3045" name="Constant_337741" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="3046" name="Constant_337734" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3047" name="Constant_337736" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3048" name="Constant_337738" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3049" name="Concat_337739" type="Concat" version="opset1">
-			<data axis="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3050" name="Broadcast_337742" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3051" name="ReadValue_330441" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.12.keypresent.12.key" variable_type="f32" variable_shape="?,32,?,80" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32" names="past_key_values.12.key">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3052" name="Constant_323457" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64" />
-			</output>
-		</layer>
-		<layer id="3053" name="Gather_323458" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I64" />
-			</input>
-			<output>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3054" name="__module.model.gpt_neox.layers.12.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3055" name="__module.model.gpt_neox.layers.12.attention/aten::size/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="3056" name="__module.model.gpt_neox.layers.12.attention/aten::size/Gather_3" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="2131" />
-			</output>
-		</layer>
-		<layer id="3057" name="__module.model.gpt_neox.layers.12.attention/aten::add_/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" names="2130,2134,2135,seq_len.49" />
-			</output>
-		</layer>
-		<layer id="3058" name="__module.model.gpt_neox.layers.12.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="3059" name="__module.model.gpt_neox.layers.12.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3060" name="__module.model.gpt_neox.layers.12.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3061" name="__module.model.gpt_neox.layers.12.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="2138,2141,cos.49">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3062" name="Constant_38269" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="3063" name="__module.model.gpt_neox.layers.12.attention/aten::index/Gather" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="2144">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3064" name="__module.model.gpt_neox.layers.12.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2145,cos.51">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3065" name="__module.model.gpt_neox.layers.12.attention/aten::mul/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2149">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3066" name="Constant_273564" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3067" name="Constant_273563" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3068" name="__module.model.gpt_neox.layers.12.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3069" name="__module.model.gpt_neox.layers.12.attention/aten::size/Constant_5" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="3070" name="__module.model.gpt_neox.layers.12.attention/aten::size/Gather_5" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="2150,2155" />
-			</output>
-		</layer>
-		<layer id="3071" name="__module.model.gpt_neox.layers.12.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" />
-			</output>
-		</layer>
-		<layer id="3072" name="__module.model.gpt_neox.layers.12.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-			</input>
-			<output>
-				<port id="1" precision="I64" names="2152,2153,2157,2158" />
-			</output>
-		</layer>
-		<layer id="3073" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Constant_1" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="3074" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3075" name="Constant_273562" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3076" name="ScatterUpdate_273565" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3077" name="Constant_273568" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3078" name="Constant_273571" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3079" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="2159,x2.49">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3080" name="Constant_299196_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3081" name="Constant_299196" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3082" name="__module.model.gpt_neox.layers.12.attention/aten::neg/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2160">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3083" name="Constant_273616" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3084" name="Constant_273621" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3085" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="2154">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3086" name="__module.model.gpt_neox.layers.12.attention/aten::cat/Concat" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2162">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3087" name="self.model.gpt_neox.layers.12.attention.rotary_emb.sin_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201162356" size="81920" />
-			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.12.attention.rotary_emb.sin_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3088" name="self.model.gpt_neox.layers.12.attention.rotary_emb.sin_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3089" name="__module.model.gpt_neox.layers.12.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3090" name="__module.model.gpt_neox.layers.12.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3091" name="__module.model.gpt_neox.layers.12.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="2139,2142,sin.49">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3092" name="Constant_38272" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="3093" name="__module.model.gpt_neox.layers.12.attention/aten::index/Gather_1" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="2147">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3094" name="__module.model.gpt_neox.layers.12.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2148,sin.51">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3095" name="__module.model.gpt_neox.layers.12.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2163">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3096" name="__module.model.gpt_neox.layers.12.attention/aten::add/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2164">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3097" name="Constant_273667" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3098" name="Constant_273670" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3099" name="Constant_273673" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3100" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="2126">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3101" name="__module.model.gpt_neox.layers.12.attention/aten::cat/Concat_1" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2182,2192,query.75,query.77">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3102" name="Constant_273679" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3103" name="Constant_273682" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3104" name="Constant_273685" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3105" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="2127,k.25">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3106" name="__module.model.gpt_neox.layers.12.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2165">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3107" name="Constant_273689" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3108" name="Constant_273688" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3109" name="__module.model.gpt_neox.layers.12.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3110" name="__module.model.gpt_neox.layers.12.attention/aten::size/Constant_7" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="3111" name="__module.model.gpt_neox.layers.12.attention/aten::size/Gather_7" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="2166,2171" />
-			</output>
-		</layer>
-		<layer id="3112" name="__module.model.gpt_neox.layers.12.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" />
-			</output>
-		</layer>
-		<layer id="3113" name="__module.model.gpt_neox.layers.12.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-			</input>
-			<output>
-				<port id="1" precision="I64" names="2168,2169,2173,2174" />
-			</output>
-		</layer>
-		<layer id="3114" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="3115" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3116" name="Constant_273687" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3117" name="ScatterUpdate_273690" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3118" name="Constant_273693" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3119" name="Constant_273696" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3120" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="2175,x2.51">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3121" name="Constant_299197_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3122" name="Constant_299197" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3123" name="__module.model.gpt_neox.layers.12.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2176">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3124" name="Constant_273741" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3125" name="Constant_273746" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3126" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="2170">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3127" name="__module.model.gpt_neox.layers.12.attention/aten::cat/Concat_2" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2178">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3128" name="__module.model.gpt_neox.layers.12.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2179">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3129" name="__module.model.gpt_neox.layers.12.attention/aten::add/Add_1" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2180">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3130" name="Constant_273791" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3131" name="Constant_273794" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3132" name="Constant_273797" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3133" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="2128">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3134" name="__module.model.gpt_neox.layers.12.attention/aten::cat/Concat_3" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2184">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3135" name="__module.model.gpt_neox.layers.12.attention/aten::to/Convert_1" type="Concat" version="opset1">
-			<data axis="-2" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="present.12.key">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3136" name="Constant_337754_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="3137" name="Constant_337754" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="3138" name="Constant_337747" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3139" name="Constant_337749" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3140" name="Constant_337751" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3141" name="Concat_337752" type="Concat" version="opset1">
-			<data axis="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3142" name="Broadcast_337755" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3143" name="ReadValue_330443" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.12.valuepresent.12.value" variable_type="f32" variable_shape="?,32,?,80" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32" names="past_key_values.12.value">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3144" name="Constant_323460" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64" />
-			</output>
-		</layer>
-		<layer id="3145" name="Gather_323461" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I64" />
-			</input>
-			<output>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3146" name="Constant_273803" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3147" name="Constant_273806" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3148" name="Constant_273809" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3149" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="2122">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3150" name="Constant_7223" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="2123">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3151" name="__module.model.gpt_neox.layers.12.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2124">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3152" name="__module.model.gpt_neox.layers.12.attention/aten::cat/Concat_5" type="Concat" version="opset1">
-			<data axis="-2" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="present.12.value">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3153" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Slice_11" type="Const" version="opset1">
-			<data element_type="boolean" shape="1, 1, 2048, 2048" offset="201244276" size="4194304" />
-			<output>
-				<port id="0" precision="BOOL" names="2201">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3154" name="Constant_273813" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3155" name="Constant_273812" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162298" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3156" name="__module.model.gpt_neox.layers.12.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3157" name="__module.model.gpt_neox.layers.12.attention/aten::size/Constant_8" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="3158" name="__module.model.gpt_neox.layers.12.attention/aten::size/Gather_8" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="2196" />
-			</output>
-		</layer>
-		<layer id="3159" name="__module.model.gpt_neox.layers.12.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3160" name="__module.model.gpt_neox.layers.12.attention/aten::size/Constant_9" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="3161" name="__module.model.gpt_neox.layers.12.attention/aten::size/Gather_9" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="2194" />
-			</output>
-		</layer>
-		<layer id="3162" name="__module.model.gpt_neox.layers.12.attention/aten::sub/Subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" names="2198,2199" />
-			</output>
-		</layer>
-		<layer id="3163" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Constant_4" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="3164" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3165" name="Constant_273811" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3166" name="ScatterUpdate_273814" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3167" name="Constant_273815" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3168" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3169" name="ScatterUpdate_273816" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3170" name="Constant_273819" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438612" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3171" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>3</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="BOOL" names="2202">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3172" name="Constant_273892" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3173" name="Constant_273893" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3174" name="Constant_273889" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3175" name="Constant_273888" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3176" name="ScatterUpdate_273894" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3177" name="Constant_273897" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3178" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="BOOL" names="2203,2204">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3179" name="__module.model.gpt_neox.layers.12.attention/aten::where/Convert_1_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="3180" name="__module.model.gpt_neox.layers.12.attention/aten::where/Convert_1" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="3181" name="__module.model.gpt_neox.layers.12.attention/aten::full/Convert" type="Const" version="opset1">
-			<data element_type="f32" shape="" offset="205438636" size="4" />
-			<output>
-				<port id="0" precision="FP32" names="2191" />
-			</output>
-		</layer>
-		<layer id="3182" name="__module.model.gpt_neox.layers.12.attention/aten::where/Select" type="Select" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="1" precision="FP32" />
-				<port id="2" precision="FP32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="2205,causal_mask.49">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3183" name="Add_290410" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3184" name="ShapeOf_290411" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3185" name="Constant_254509" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3186" name="__module.model.gpt_neox.layers.12.attention/aten::size/Constant_10" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="3187" name="__module.model.gpt_neox.layers.12.attention/aten::size/Gather_10" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="2189">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3188" name="__module.model.gpt_neox.layers.12.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1">
-			<data axis="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64" names="2206">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3189" name="__module.model.gpt_neox.layers.12.attention/aten::expand/Abs" type="Abs" version="opset1">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3190" name="Maximum_290412" type="Maximum" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3191" name="__module.model.gpt_neox.layers.12.attention/aten::add/Add_2" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2208">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3192" name="__module.model.gpt_neox.layers.12.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
-			<data causal="false" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="2209,2210,sdpa_result.25,tensor.49">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3193" name="Constant_7664" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="2211">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3194" name="__module.model.gpt_neox.layers.12.attention/aten::permute/Transpose_3" type="Transpose" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2212">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3195" name="__module.model.gpt_neox.layers.12.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438650" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3196" name="__module.model.gpt_neox.layers.12.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
-			<data special_zero="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>3</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2217">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3197" name="self.model.gpt_neox.layers.12.attention.dense.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 2560" offset="1150785234" size="6553600" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3198" name="Convert_21914" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3199" name="self.model.gpt_neox.layers.12.attention.dense.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="1157338834" size="2560" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3200" name="Convert_21917" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3201" name="self.model.gpt_neox.layers.12.attention.dense.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3202" name="self.model.gpt_neox.layers.12.attention.dense.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="1157341394" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3203" name="self.model.gpt_neox.layers.12.attention.dense.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3204" name="self.model.gpt_neox.layers.12.attention.dense.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3205" name="__module.model.gpt_neox.layers.12.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3206" name="Constant_299198_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1157346514" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3207" name="Constant_299198" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3208" name="__module.model.gpt_neox.layers.12.attention.dense/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2220,2222,input.51">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3209" name="__module.model.gpt_neox.layers.12/aten::add/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2240">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3210" name="__module.model.gpt_neox.layers.12/aten::add/Add_1" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2241,2243,hidden_states.27">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3211" name="__module.model.gpt_neox.layers.13.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867670" size="4" />
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3212" name="__module.model.gpt_neox.layers.13.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
-			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3213" name="Constant_299199_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1157351634" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3214" name="Constant_299199" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3215" name="__module.model.gpt_neox.layers.13.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3216" name="Constant_299200_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1157356754" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3217" name="Constant_299200" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3218" name="__module.model.gpt_neox.layers.13.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2378">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3219" name="self.model.gpt_neox.layers.13.mlp.dense_h_to_4h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 2560" offset="1157361874" size="26214400" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3220" name="Convert_21964" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3221" name="self.model.gpt_neox.layers.13.mlp.dense_h_to_4h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 1" offset="1183576274" size="10240" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3222" name="Convert_21967" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3223" name="self.model.gpt_neox.layers.13.mlp.dense_h_to_4h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3224" name="self.model.gpt_neox.layers.13.mlp.dense_h_to_4h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="10240, 1" offset="1183586514" size="20480" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3225" name="self.model.gpt_neox.layers.13.mlp.dense_h_to_4h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3226" name="self.model.gpt_neox.layers.13.mlp.dense_h_to_4h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3227" name="__module.model.gpt_neox.layers.13.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3228" name="Constant_299201_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 10240" offset="1183606994" size="20480" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3229" name="Constant_299201" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3230" name="__module.model.gpt_neox.layers.13.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2383">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3231" name="__module.model.gpt_neox.layers.13.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
-			<data approximation_mode="ERF" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32" names="2384">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3232" name="self.model.gpt_neox.layers.13.mlp.dense_4h_to_h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 10240" offset="1183627474" size="26214400" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3233" name="Convert_21974" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3234" name="self.model.gpt_neox.layers.13.mlp.dense_4h_to_h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="1209841874" size="2560" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3235" name="Convert_21977" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3236" name="self.model.gpt_neox.layers.13.mlp.dense_4h_to_h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3237" name="self.model.gpt_neox.layers.13.mlp.dense_4h_to_h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="1209844434" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3238" name="self.model.gpt_neox.layers.13.mlp.dense_4h_to_h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3239" name="self.model.gpt_neox.layers.13.mlp.dense_4h_to_h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3240" name="__module.model.gpt_neox.layers.13.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3241" name="Constant_299202_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1209849554" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3242" name="Constant_299202" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3243" name="__module.model.gpt_neox.layers.13.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2387,input.57">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3244" name="Constant_299203_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1209854674" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3245" name="Constant_299203" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3246" name="__module.model.gpt_neox.layers.13.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3247" name="Constant_299204_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1209859794" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3248" name="Constant_299204" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3249" name="__module.model.gpt_neox.layers.13.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2253">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3250" name="self.model.gpt_neox.layers.13.attention.query_key_value.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 2560" offset="1209864914" size="19660800" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3251" name="Convert_21944" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3252" name="self.model.gpt_neox.layers.13.attention.query_key_value.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 1" offset="1229525714" size="7680" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3253" name="Convert_21947" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3254" name="self.model.gpt_neox.layers.13.attention.query_key_value.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3255" name="self.model.gpt_neox.layers.13.attention.query_key_value.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="7680, 1" offset="1229533394" size="15360" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3256" name="self.model.gpt_neox.layers.13.attention.query_key_value.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3257" name="self.model.gpt_neox.layers.13.attention.query_key_value.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3258" name="__module.model.gpt_neox.layers.13.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3259" name="Constant_299205_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 7680" offset="1229548754" size="15360" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3260" name="Constant_299205" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3261" name="__module.model.gpt_neox.layers.13.attention.query_key_value/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2260,qkv.53">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3262" name="__module.model.gpt_neox.layers.13.attention/prim::ListConstruct/Concat" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080154" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3263" name="__module.model.gpt_neox.layers.13.attention/aten::view/Reshape" type="Reshape" version="opset1">
-			<data special_zero="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2264,qkv.55">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3264" name="Constant_273904" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3265" name="Constant_273907" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3266" name="Constant_273910" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3267" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="2265">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3268" name="Constant_7776" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="2266">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3269" name="__module.model.gpt_neox.layers.13.attention/aten::permute/Transpose" type="Transpose" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2267,query.79">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3270" name="Constant_273916" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3271" name="Constant_273919" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3272" name="Constant_273922" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3273" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="2274,q.27">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3274" name="self.model.gpt_neox.layers.13.attention.rotary_emb.cos_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201080346" size="81920" />
-			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.13.attention.rotary_emb.cos_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3275" name="self.model.gpt_neox.layers.13.attention.rotary_emb.cos_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3276" name="__module.model.gpt_neox.layers.13.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3277" name="Constant_273928" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3278" name="Constant_273931" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3279" name="Constant_273934" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3280" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="2268">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3281" name="Constant_7790" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="2269">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3282" name="__module.model.gpt_neox.layers.13.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2270,key.79">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3283" name="__module.model.gpt_neox.layers.13.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3284" name="__module.model.gpt_neox.layers.13.attention/aten::size/Constant_2" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="3285" name="__module.model.gpt_neox.layers.13.attention/aten::size/Gather_2" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="2278" />
-			</output>
-		</layer>
-		<layer id="3286" name="Constant_337767_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="3287" name="Constant_337767" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="3288" name="Constant_337760" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3289" name="Constant_337762" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3290" name="Constant_337764" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3291" name="Concat_337765" type="Concat" version="opset1">
-			<data axis="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3292" name="Broadcast_337768" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3293" name="ReadValue_330445" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.13.keypresent.13.key" variable_type="f32" variable_shape="?,32,?,80" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32" names="past_key_values.13.key">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3294" name="Constant_323463" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64" />
-			</output>
-		</layer>
-		<layer id="3295" name="Gather_323464" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I64" />
-			</input>
-			<output>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3296" name="__module.model.gpt_neox.layers.13.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3297" name="__module.model.gpt_neox.layers.13.attention/aten::size/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="3298" name="__module.model.gpt_neox.layers.13.attention/aten::size/Gather_3" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="2280" />
-			</output>
-		</layer>
-		<layer id="3299" name="__module.model.gpt_neox.layers.13.attention/aten::add_/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" names="2279,2283,2284,seq_len.53" />
-			</output>
-		</layer>
-		<layer id="3300" name="__module.model.gpt_neox.layers.13.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="3301" name="__module.model.gpt_neox.layers.13.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3302" name="__module.model.gpt_neox.layers.13.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3303" name="__module.model.gpt_neox.layers.13.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="2287,2290,cos.53">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3304" name="Constant_38275" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="3305" name="__module.model.gpt_neox.layers.13.attention/aten::index/Gather" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="2293">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3306" name="__module.model.gpt_neox.layers.13.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2294,cos.55">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3307" name="__module.model.gpt_neox.layers.13.attention/aten::mul/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2298">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3308" name="Constant_273939" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3309" name="Constant_273938" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3310" name="__module.model.gpt_neox.layers.13.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3311" name="__module.model.gpt_neox.layers.13.attention/aten::size/Constant_5" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="3312" name="__module.model.gpt_neox.layers.13.attention/aten::size/Gather_5" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="2299,2304" />
-			</output>
-		</layer>
-		<layer id="3313" name="__module.model.gpt_neox.layers.13.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" />
-			</output>
-		</layer>
-		<layer id="3314" name="__module.model.gpt_neox.layers.13.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-			</input>
-			<output>
-				<port id="1" precision="I64" names="2301,2302,2306,2307" />
-			</output>
-		</layer>
-		<layer id="3315" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Constant_1" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="3316" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3317" name="Constant_273937" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3318" name="ScatterUpdate_273940" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3319" name="Constant_273943" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3320" name="Constant_273946" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3321" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="2308,x2.53">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3322" name="Constant_299206_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3323" name="Constant_299206" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3324" name="__module.model.gpt_neox.layers.13.attention/aten::neg/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2309">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3325" name="Constant_273991" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3326" name="Constant_273996" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3327" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="2303">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3328" name="__module.model.gpt_neox.layers.13.attention/aten::cat/Concat" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2311">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3329" name="self.model.gpt_neox.layers.13.attention.rotary_emb.sin_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201162356" size="81920" />
-			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.13.attention.rotary_emb.sin_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3330" name="self.model.gpt_neox.layers.13.attention.rotary_emb.sin_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3331" name="__module.model.gpt_neox.layers.13.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3332" name="__module.model.gpt_neox.layers.13.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3333" name="__module.model.gpt_neox.layers.13.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="2288,2291,sin.53">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3334" name="Constant_38278" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="3335" name="__module.model.gpt_neox.layers.13.attention/aten::index/Gather_1" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="2296">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3336" name="__module.model.gpt_neox.layers.13.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2297,sin.55">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3337" name="__module.model.gpt_neox.layers.13.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2312">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3338" name="__module.model.gpt_neox.layers.13.attention/aten::add/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2313">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3339" name="Constant_274042" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3340" name="Constant_274045" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3341" name="Constant_274048" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3342" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="2275">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3343" name="__module.model.gpt_neox.layers.13.attention/aten::cat/Concat_1" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2331,2341,query.81,query.83">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3344" name="Constant_274054" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3345" name="Constant_274057" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3346" name="Constant_274060" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3347" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="2276,k.27">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3348" name="__module.model.gpt_neox.layers.13.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2314">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3349" name="Constant_274064" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3350" name="Constant_274063" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3351" name="__module.model.gpt_neox.layers.13.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3352" name="__module.model.gpt_neox.layers.13.attention/aten::size/Constant_7" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="3353" name="__module.model.gpt_neox.layers.13.attention/aten::size/Gather_7" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="2315,2320" />
-			</output>
-		</layer>
-		<layer id="3354" name="__module.model.gpt_neox.layers.13.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" />
-			</output>
-		</layer>
-		<layer id="3355" name="__module.model.gpt_neox.layers.13.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-			</input>
-			<output>
-				<port id="1" precision="I64" names="2317,2318,2322,2323" />
-			</output>
-		</layer>
-		<layer id="3356" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="3357" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3358" name="Constant_274062" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3359" name="ScatterUpdate_274065" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3360" name="Constant_274068" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3361" name="Constant_274071" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3362" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="2324,x2.55">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3363" name="Constant_299207_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3364" name="Constant_299207" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3365" name="__module.model.gpt_neox.layers.13.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2325">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3366" name="Constant_274116" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3367" name="Constant_274121" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3368" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="2319">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3369" name="__module.model.gpt_neox.layers.13.attention/aten::cat/Concat_2" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2327">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3370" name="__module.model.gpt_neox.layers.13.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2328">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3371" name="__module.model.gpt_neox.layers.13.attention/aten::add/Add_1" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2329">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3372" name="Constant_274166" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3373" name="Constant_274169" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3374" name="Constant_274172" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3375" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="2277">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3376" name="__module.model.gpt_neox.layers.13.attention/aten::cat/Concat_3" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2333">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3377" name="__module.model.gpt_neox.layers.13.attention/aten::to/Convert_1" type="Concat" version="opset1">
-			<data axis="-2" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="present.13.key">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3378" name="Constant_337780_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="3379" name="Constant_337780" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="3380" name="Constant_337773" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3381" name="Constant_337775" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3382" name="Constant_337777" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3383" name="Concat_337778" type="Concat" version="opset1">
-			<data axis="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3384" name="Broadcast_337781" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3385" name="ReadValue_330447" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.13.valuepresent.13.value" variable_type="f32" variable_shape="?,32,?,80" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32" names="past_key_values.13.value">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3386" name="Constant_323466" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64" />
-			</output>
-		</layer>
-		<layer id="3387" name="Gather_323467" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I64" />
-			</input>
-			<output>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3388" name="Constant_274178" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3389" name="Constant_274181" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3390" name="Constant_274184" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="I32">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3391" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="907" name="__module.model.gpt_neox.layers.4.attention/aten::transpose/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="3" precision="I64">
+				<port id="1" precision="I32">
 					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2271">
+				<port id="2" precision="FP32" names="906">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -51771,15 +13592,16 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="3392" name="Constant_7804" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="908" name="Constant_40988" type="Const" version="opset1">
+			<data element_type="i64" shape="3" offset="201116216" size="24" />
 			<output>
-				<port id="0" precision="I64" names="2272">
-					<dim>4</dim>
+				<port id="0" precision="I64">
+					<dim>3</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3393" name="__module.model.gpt_neox.layers.13.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
+		<layer id="909" name="__module.model.gpt_neox.layers.4.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
+			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -51788,807 +13610,864 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>3</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2273">
+				<port id="2" precision="FP32" names="909">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3394" name="__module.model.gpt_neox.layers.13.attention/aten::cat/Concat_5" type="Concat" version="opset1">
-			<data axis="-2" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+		<layer id="910" name="self.model.gpt_neox.layers.4.attention.dense.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 2560" offset="516395600" size="6553600" />
+			<output>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+			</output>
+		</layer>
+		<layer id="911" name="Convert_75220" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="present.13.value">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3395" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Slice_11" type="Const" version="opset1">
-			<data element_type="boolean" shape="1, 1, 2048, 2048" offset="201244276" size="4194304" />
+		<layer id="912" name="self.model.gpt_neox.layers.4.attention.dense.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="522949200" size="2560" />
 			<output>
-				<port id="0" precision="BOOL" names="2350">
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="913" name="Convert_75223" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
 					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3396" name="Constant_274188" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="914" name="self.model.gpt_neox.layers.4.attention.dense.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3397" name="Constant_274187" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162298" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="915" name="self.model.gpt_neox.layers.4.attention.dense.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="522951760" size="5120" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3398" name="__module.model.gpt_neox.layers.13.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="916" name="self.model.gpt_neox.layers.4.attention.dense.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3399" name="__module.model.gpt_neox.layers.13.attention/aten::size/Constant_8" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="3400" name="__module.model.gpt_neox.layers.13.attention/aten::size/Gather_8" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="917" name="self.model.gpt_neox.layers.4.attention.dense.weight/fq_weights_1/convert" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="2345" />
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3401" name="__module.model.gpt_neox.layers.13.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="918" name="__module.model.gpt_neox.layers.4.attention.dense/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3402" name="__module.model.gpt_neox.layers.13.attention/aten::size/Constant_9" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="919" name="Constant_37327" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="522956880" size="10240" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3403" name="__module.model.gpt_neox.layers.13.attention/aten::size/Gather_9" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="920" name="__module.model.gpt_neox.layers.4.attention.dense/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="2343" />
+				<port id="2" precision="FP32" names="912,914,input.17">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3404" name="__module.model.gpt_neox.layers.13.attention/aten::sub/Subtract" type="Subtract" version="opset1">
+		<layer id="921" name="__module.model.gpt_neox.layers.4/aten::add/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64" names="2347,2348" />
-			</output>
-		</layer>
-		<layer id="3405" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Constant_4" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
+				<port id="2" precision="FP32" names="932">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3406" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1">
+		<layer id="922" name="__module.model.gpt_neox.layers.4/aten::add/Add_1" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="933,935,hidden_states.23">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3407" name="Constant_274186" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="923" name="__module.model.gpt_neox.layers.5.input_layernorm/aten::layer_norm/Multiply" type="Const" version="opset1">
+			<data element_type="i32" shape="1" offset="128867668" size="4" />
 			<output>
 				<port id="0" precision="I32">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3408" name="ScatterUpdate_274189" type="ScatterUpdate" version="opset3">
+		<layer id="924" name="__module.model.gpt_neox.layers.5.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
+			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="1" precision="I32">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3409" name="Constant_274190" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3410" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="925" name="Constant_37328" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="522967120" size="10240" />
 			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3411" name="ScatterUpdate_274191" type="ScatterUpdate" version="opset3">
+		<layer id="926" name="__module.model.gpt_neox.layers.5.post_attention_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="2" precision="I64">
+				<port id="1" precision="FP32">
 					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
 					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3412" name="Constant_274194" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438612" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="927" name="Constant_37329" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="522977360" size="10240" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3413" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="928" name="__module.model.gpt_neox.layers.5.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="BOOL">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>3</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="BOOL" names="2351">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="1033">
+					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>2048</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3414" name="Constant_274267" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="929" name="self.model.gpt_neox.layers.5.mlp.dense_h_to_4h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 2560" offset="522987600" size="26214400" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3415" name="Constant_274268" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="930" name="Convert_74284" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3416" name="Constant_274264" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="931" name="self.model.gpt_neox.layers.5.mlp.dense_h_to_4h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 1" offset="549202000" size="10240" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="U8">
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3417" name="Constant_274263" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="932" name="Convert_74287" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I32">
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3418" name="ScatterUpdate_274269" type="ScatterUpdate" version="opset3">
+		<layer id="933" name="self.model.gpt_neox.layers.5.mlp.dense_h_to_4h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="934" name="self.model.gpt_neox.layers.5.mlp.dense_h_to_4h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="10240, 1" offset="549212240" size="20480" />
+			<output>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+			</output>
+		</layer>
+		<layer id="935" name="self.model.gpt_neox.layers.5.mlp.dense_h_to_4h.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3419" name="Constant_274272" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
+		<layer id="936" name="__module.model.gpt_neox.layers.5.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<rt_info>
-				<attribute name="precise" version="0" />
+				<attribute name="decompression" version="0" />
 			</rt_info>
+			<input>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3420" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="937" name="__module.model.gpt_neox.layers.5.mlp.dense_h_to_4h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="BOOL" names="2352,2353">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3421" name="__module.model.gpt_neox.layers.13.attention/aten::where/Convert_1_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
+		<layer id="938" name="self.model.gpt_neox.layers.5.mlp.dense_h_to_4h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="10240" offset="549232720" size="20480" />
 			<output>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.5.mlp.dense_h_to_4h.bias">
+					<dim>10240</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3422" name="__module.model.gpt_neox.layers.13.attention/aten::where/Convert_1" type="Convert" version="opset1">
+		<layer id="939" name="__module.model.gpt_neox.layers.5.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
 			</rt_info>
 			<input>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="BF16">
+					<dim>10240</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="3423" name="__module.model.gpt_neox.layers.13.attention/aten::full/Convert" type="Const" version="opset1">
-			<data element_type="f32" shape="" offset="205438636" size="4" />
-			<output>
-				<port id="0" precision="FP32" names="2340" />
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3424" name="__module.model.gpt_neox.layers.13.attention/aten::where/Select" type="Select" version="opset1">
+		<layer id="940" name="__module.model.gpt_neox.layers.5.mlp.dense_h_to_4h/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>10240</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
 				</port>
-				<port id="1" precision="FP32" />
-				<port id="2" precision="FP32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="2354,causal_mask.53">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="1038">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3425" name="Add_290416" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="941" name="__module.model.gpt_neox.layers.5.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
+			<data approximation_mode="ERF" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="1" precision="FP32">
 					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
+				<port id="1" precision="FP32" names="1039">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3426" name="ShapeOf_290417" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="942" name="self.model.gpt_neox.layers.5.mlp.dense_4h_to_h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 10240" offset="549253200" size="26214400" />
+			<output>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="943" name="Convert_74294" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3427" name="Constant_254524" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="944" name="self.model.gpt_neox.layers.5.mlp.dense_4h_to_h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="575467600" size="2560" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="U8">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3428" name="__module.model.gpt_neox.layers.13.attention/aten::size/Constant_10" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="945" name="Convert_74297" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I32" />
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3429" name="__module.model.gpt_neox.layers.13.attention/aten::size/Gather_10" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="946" name="self.model.gpt_neox.layers.5.mlp.dense_4h_to_h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I32" />
-			</input>
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="947" name="self.model.gpt_neox.layers.5.mlp.dense_4h_to_h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="575470160" size="5120" />
 			<output>
-				<port id="3" precision="I64" names="2338">
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3430" name="__module.model.gpt_neox.layers.13.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="948" name="self.model.gpt_neox.layers.5.mlp.dense_4h_to_h.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="3" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64" names="2355">
-					<dim>4</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3431" name="__module.model.gpt_neox.layers.13.attention/aten::expand/Abs" type="Abs" version="opset1">
+		<layer id="949" name="__module.model.gpt_neox.layers.5.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<rt_info>
+				<attribute name="decompression" version="0" />
+			</rt_info>
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3432" name="Maximum_290418" type="Maximum" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="950" name="__module.model.gpt_neox.layers.5.mlp.dense_4h_to_h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3433" name="__module.model.gpt_neox.layers.13.attention/aten::add/Add_2" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
+		<layer id="951" name="self.model.gpt_neox.layers.5.mlp.dense_4h_to_h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="2560" offset="575475280" size="5120" />
+			<output>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.5.mlp.dense_4h_to_h.bias">
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+			</output>
+		</layer>
+		<layer id="952" name="__module.model.gpt_neox.layers.5.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<rt_info>
+				<attribute name="decompression" version="0" />
+			</rt_info>
+			<input>
+				<port id="0" precision="BF16">
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2357">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3434" name="__module.model.gpt_neox.layers.13.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
-			<data causal="false" />
+		<layer id="953" name="__module.model.gpt_neox.layers.5.mlp.dense_4h_to_h/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2358,2359,sdpa_result.27,tensor.53">
+				<port id="2" precision="FP32" names="1042,input.23">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3435" name="Constant_8245" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="954" name="Constant_37330" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="575480400" size="10240" />
 			<output>
-				<port id="0" precision="I64" names="2360">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3436" name="__module.model.gpt_neox.layers.13.attention/aten::permute/Transpose_3" type="Transpose" version="opset1">
+		<layer id="955" name="__module.model.gpt_neox.layers.5.input_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2361">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3437" name="__module.model.gpt_neox.layers.13.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438650" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="956" name="Constant_37331" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="575490640" size="10240" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3438" name="__module.model.gpt_neox.layers.13.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
-			<data special_zero="true" />
+		<layer id="957" name="__module.model.gpt_neox.layers.5.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>3</dim>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2366">
+				<port id="2" precision="FP32" names="945,hidden_states.21">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3439" name="self.model.gpt_neox.layers.13.attention.dense.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 2560" offset="1229564114" size="6553600" />
+		<layer id="958" name="self.model.gpt_neox.layers.5.attention.query_key_value.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 2560" offset="575500880" size="19660800" />
 			<output>
 				<port id="0" precision="U8">
-					<dim>2560</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3440" name="Convert_21954" type="Convert" version="opset1">
+		<layer id="959" name="Convert_74879" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
-					<dim>2560</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP16">
-					<dim>2560</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3441" name="self.model.gpt_neox.layers.13.attention.dense.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="1236117714" size="2560" />
+		<layer id="960" name="self.model.gpt_neox.layers.5.attention.query_key_value.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 1" offset="595161680" size="7680" />
 			<output>
 				<port id="0" precision="U8">
-					<dim>2560</dim>
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3442" name="Convert_21957" type="Convert" version="opset1">
+		<layer id="961" name="Convert_74882" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
-					<dim>2560</dim>
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP16">
-					<dim>2560</dim>
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3443" name="self.model.gpt_neox.layers.13.attention.dense.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="962" name="self.model.gpt_neox.layers.5.attention.query_key_value.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
-					<dim>2560</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP16">
-					<dim>2560</dim>
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
 				<port id="2" precision="FP16">
-					<dim>2560</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3444" name="self.model.gpt_neox.layers.13.attention.dense.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="1236120274" size="5120" />
+		<layer id="963" name="self.model.gpt_neox.layers.5.attention.query_key_value.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="7680, 1" offset="595169360" size="15360" />
 			<output>
 				<port id="0" precision="FP16">
-					<dim>2560</dim>
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3445" name="self.model.gpt_neox.layers.13.attention.dense.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="964" name="self.model.gpt_neox.layers.5.attention.query_key_value.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
-					<dim>2560</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP16">
-					<dim>2560</dim>
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
 				<port id="2" precision="FP16">
-					<dim>2560</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3446" name="self.model.gpt_neox.layers.13.attention.dense.weight" type="Convert" version="opset1">
+		<layer id="965" name="self.model.gpt_neox.layers.5.attention.query_key_value.weight/fq_weights_1/convert" type="Convert" version="opset1">
 			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
 			<input>
 				<port id="0" precision="FP16">
-					<dim>2560</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP32">
-					<dim>2560</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3447" name="__module.model.gpt_neox.layers.13.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="966" name="__module.model.gpt_neox.layers.5.attention.query_key_value/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -52597,7 +14476,7 @@
 					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>2560</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
@@ -52605,1032 +14484,1126 @@
 				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3448" name="Constant_299208_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1236125394" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3449" name="Constant_299208" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="967" name="Constant_37332" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 7680" offset="595184720" size="30720" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3450" name="__module.model.gpt_neox.layers.13.attention.dense/aten::linear/Add" type="Add" version="opset1">
+		<layer id="968" name="__module.model.gpt_neox.layers.5.attention.query_key_value/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>7680</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>7680</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2369,2371,input.55">
+				<port id="2" precision="FP32" names="952,qkv.21">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3451" name="__module.model.gpt_neox.layers.13/aten::add/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="969" name="__module.model.gpt_neox.layers.5.attention/prim::ListConstruct" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201115992" size="32" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="970" name="__module.model.gpt_neox.layers.5.attention/aten::view/Reshape" type="Reshape" version="opset1">
+			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>7680</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2389">
+				<port id="2" precision="FP32" names="956,qkv.23">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3452" name="__module.model.gpt_neox.layers.13/aten::add/Add_1" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="971" name="Constant_3866" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="972" name="Constant_3868" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="973" name="Constant_3870" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="974" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Slice" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2390,2392,hidden_states.29">
+				<port id="5" precision="FP32" names="957">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3453" name="__module.model.gpt_neox.layers.14.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867670" size="4" />
+		<layer id="975" name="Constant_3879" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
+				<port id="0" precision="I64" names="958">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3454" name="__module.model.gpt_neox.layers.14.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
-			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
+		<layer id="976" name="__module.model.gpt_neox.layers.5.attention/aten::permute/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="I32">
-					<dim>1</dim>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="959,query.11">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3455" name="Constant_299209_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1236130514" size="5120" />
+		<layer id="977" name="Constant_3920" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="FP16">
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="978" name="Constant_3922" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3456" name="Constant_299209" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="979" name="Constant_3924" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="980" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Slice_1" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
+				<port id="3" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="5" precision="FP32" names="966,q.11">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3457" name="__module.model.gpt_neox.layers.14.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
+		<layer id="981" name="__module.model.gpt_neox.layers.5.attention/aten::mul/Multiply" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
+					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="972">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3458" name="Constant_299210_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1236135634" size="5120" />
+		<layer id="982" name="Constant_4078" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
-				<port id="0" precision="FP16">
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="983" name="Constant_4080" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3459" name="Constant_299210" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="984" name="Constant_4082" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="985" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Slice_2" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
+				<port id="5" precision="FP32" names="982,x2.21">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="986" name="Constant_37333" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3460" name="__module.model.gpt_neox.layers.14.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+		<layer id="987" name="__module.model.gpt_neox.layers.5.attention/aten::neg/Multiply" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2527">
+				<port id="2" precision="FP32" names="983">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3461" name="self.model.gpt_neox.layers.14.mlp.dense_h_to_4h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 2560" offset="1236140754" size="26214400" />
+		<layer id="988" name="Constant_3975" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3462" name="Convert_22004" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="989" name="Constant_4019" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3463" name="self.model.gpt_neox.layers.14.mlp.dense_h_to_4h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 1" offset="1262355154" size="10240" />
+		<layer id="990" name="Constant_4021" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3464" name="Convert_22007" type="Convert" version="opset1">
-			<data destination_type="f16" />
+		<layer id="991" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Slice_3" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
+				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="3465" name="self.model.gpt_neox.layers.14.mlp.dense_h_to_4h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="3" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3466" name="self.model.gpt_neox.layers.14.mlp.dense_h_to_4h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="10240, 1" offset="1262365394" size="20480" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
+				<port id="5" precision="FP32" names="977">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3467" name="self.model.gpt_neox.layers.14.mlp.dense_h_to_4h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="992" name="__module.model.gpt_neox.layers.5.attention/aten::cat/Concat" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32" names="985">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3468" name="self.model.gpt_neox.layers.14.mlp.dense_h_to_4h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="993" name="__module.model.gpt_neox.layers.5.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32" names="986">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3469" name="__module.model.gpt_neox.layers.14.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
+		<layer id="994" name="__module.model.gpt_neox.layers.5.attention/aten::add/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>10240</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="987">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3470" name="Constant_299211_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 10240" offset="1262385874" size="20480" />
+		<layer id="995" name="Constant_3930" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3471" name="Constant_299211" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+		<layer id="996" name="Constant_3932" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>10240</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="997" name="Constant_3934" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3472" name="__module.model.gpt_neox.layers.14.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="998" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Slice_4" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2532">
+				<port id="5" precision="FP32" names="967">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>60</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3473" name="__module.model.gpt_neox.layers.14.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
-			<data approximation_mode="ERF" />
+		<layer id="999" name="__module.model.gpt_neox.layers.5.attention/aten::cat/Concat_1" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="2533">
+				<port id="2" precision="FP32" names="1005">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3474" name="self.model.gpt_neox.layers.14.mlp.dense_4h_to_h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 10240" offset="1262406354" size="26214400" />
+		<layer id="1000" name="Constant_52671" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
+				<port id="0" precision="FP32" />
 			</output>
 		</layer>
-		<layer id="3475" name="Convert_22014" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
+		<layer id="1001" name="Constant_52664" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3476" name="self.model.gpt_neox.layers.14.mlp.dense_4h_to_h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="1288620754" size="2560" />
+		<layer id="1002" name="Constant_52666" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3477" name="Convert_22017" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
+		<layer id="1003" name="Constant_52668" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3478" name="self.model.gpt_neox.layers.14.mlp.dense_4h_to_h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1004" name="Concat_52669" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="2" precision="I64">
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="3479" name="self.model.gpt_neox.layers.14.mlp.dense_4h_to_h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="1288623314" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
+			</input>
+			<output>
+				<port id="4" precision="I64">
+					<dim>4</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3480" name="self.model.gpt_neox.layers.14.mlp.dense_4h_to_h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1005" name="Broadcast_52672" type="Broadcast" version="opset3">
+			<data mode="numpy" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3481" name="self.model.gpt_neox.layers.14.mlp.dense_4h_to_h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="1006" name="ReadValue_50291" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.5.keypresent.5.key" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="1" precision="FP32" names="past_key_values.5.key">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3482" name="__module.model.gpt_neox.layers.14.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
+		<layer id="1007" name="Constant_47808" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64" />
+			</output>
+		</layer>
+		<layer id="1008" name="Gather_47809" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="1" precision="I32">
+					<dim>-1</dim>
 				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="3" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3483" name="Constant_299212_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1288628434" size="5120" />
+		<layer id="1009" name="Constant_3884" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3484" name="Constant_299212" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+		<layer id="1010" name="Constant_3886" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="1011" name="Constant_3888" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3485" name="__module.model.gpt_neox.layers.14.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1012" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Slice_5" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
-				<port id="1" precision="FP32">
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2536,input.61">
+				<port id="5" precision="FP32" names="960">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3486" name="Constant_299213_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1288633554" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3487" name="Constant_299213" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="1013" name="Constant_3897" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64" names="961">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3488" name="__module.model.gpt_neox.layers.14.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1014" name="__module.model.gpt_neox.layers.5.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="962,key.21">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3489" name="Constant_299214_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1288638674" size="5120" />
+		<layer id="1015" name="Constant_3940" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="FP16">
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1016" name="Constant_3942" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1017" name="Constant_3944" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3490" name="Constant_299214" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="1018" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Slice_6" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
+				<port id="3" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="5" precision="FP32" names="968,k.11">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3491" name="__module.model.gpt_neox.layers.14.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+		<layer id="1019" name="__module.model.gpt_neox.layers.5.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
+					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2402">
+				<port id="2" precision="FP32" names="988">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3492" name="self.model.gpt_neox.layers.14.attention.query_key_value.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 2560" offset="1288643794" size="19660800" />
+		<layer id="1020" name="Constant_4221" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3493" name="Convert_21984" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="1021" name="Constant_4223" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3494" name="self.model.gpt_neox.layers.14.attention.query_key_value.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 1" offset="1308304594" size="7680" />
+		<layer id="1022" name="Constant_4225" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3495" name="Convert_21987" type="Convert" version="opset1">
-			<data destination_type="f16" />
+		<layer id="1023" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Slice_7" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
+				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="3496" name="self.model.gpt_neox.layers.14.attention.query_key_value.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="3" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="5" precision="FP32" names="998,x2.23">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3497" name="self.model.gpt_neox.layers.14.attention.query_key_value.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="7680, 1" offset="1308312274" size="15360" />
+		<layer id="1024" name="Constant_37334" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3498" name="self.model.gpt_neox.layers.14.attention.query_key_value.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="1025" name="__module.model.gpt_neox.layers.5.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32" names="999">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3499" name="self.model.gpt_neox.layers.14.attention.query_key_value.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+		<layer id="1026" name="Constant_4118" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="1027" name="Constant_4162" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3500" name="__module.model.gpt_neox.layers.14.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
+		<layer id="1028" name="Constant_4164" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1029" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Slice_8" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="5" precision="FP32" names="993">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3501" name="Constant_299215_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 7680" offset="1308327634" size="15360" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3502" name="Constant_299215" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="1030" name="__module.model.gpt_neox.layers.5.attention/aten::cat/Concat_2" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
+				<port id="2" precision="FP32" names="1001">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3503" name="__module.model.gpt_neox.layers.14.attention.query_key_value/aten::linear/Add" type="Add" version="opset1">
+		<layer id="1031" name="__module.model.gpt_neox.layers.5.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
+					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2409,qkv.57">
+				<port id="2" precision="FP32" names="1002">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3504" name="__module.model.gpt_neox.layers.14.attention/prim::ListConstruct/Concat" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080154" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3505" name="__module.model.gpt_neox.layers.14.attention/aten::view/Reshape" type="Reshape" version="opset1">
-			<data special_zero="true" />
+		<layer id="1032" name="__module.model.gpt_neox.layers.5.attention/aten::add/Add_1" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2413,qkv.59">
-					<dim>-1</dim>
+				<port id="2" precision="FP32" names="1003">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>240</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3506" name="Constant_274279" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1033" name="Constant_3950" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3507" name="Constant_274282" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1034" name="Constant_3952" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3508" name="Constant_274285" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1035" name="Constant_3954" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3509" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1036" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Slice_9" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>240</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2414">
-					<dim>-1</dim>
+				<port id="5" precision="FP32" names="969">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3510" name="Constant_8357" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="1037" name="__module.model.gpt_neox.layers.5.attention/aten::cat/Concat_3" type="Concat" version="opset1">
+			<data axis="-1" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64" names="2415">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="1007">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3511" name="__module.model.gpt_neox.layers.14.attention/aten::permute/Transpose" type="Transpose" version="opset1">
+		<layer id="1038" name="__module.model.gpt_neox.layers.5.attention/aten::cat/Concat_4" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2416,query.85">
+				<port id="2" precision="FP32" names="present.5.key">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -53638,140 +15611,148 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="3512" name="Constant_274291" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1039" name="Constant_52683" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
+			<output>
+				<port id="0" precision="FP32" />
+			</output>
+		</layer>
+		<layer id="1040" name="Constant_52676" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3513" name="Constant_274294" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1041" name="Constant_52678" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3514" name="Constant_274297" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1042" name="Constant_52680" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3515" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1043" name="Concat_52681" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2423,q.29">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="4" precision="I64">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3516" name="self.model.gpt_neox.layers.14.attention.rotary_emb.cos_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201080346" size="81920" />
+		<layer id="1044" name="Broadcast_52684" type="Broadcast" version="opset3">
+			<data mode="numpy" />
+			<input>
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
+					<dim>4</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.14.attention.rotary_emb.cos_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3517" name="self.model.gpt_neox.layers.14.attention.rotary_emb.cos_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="1045" name="ReadValue_50293" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.5.valuepresent.5.value" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="1" precision="FP32" names="past_key_values.5.value">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3518" name="__module.model.gpt_neox.layers.14.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1046" name="Constant_47811" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="I64" />
+			</output>
+		</layer>
+		<layer id="1047" name="Gather_47812" type="Gather" version="opset8">
+			<data batch_dims="0" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I32">
+					<dim>-1</dim>
+				</port>
+				<port id="2" precision="I64" />
+			</input>
+			<output>
+				<port id="3" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3519" name="Constant_274303" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1048" name="Constant_3902" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3520" name="Constant_274306" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1049" name="Constant_3904" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3521" name="Constant_274309" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1050" name="Constant_3906" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3522" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1051" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Slice_10" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -53780,17 +15761,20 @@
 					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2417">
+				<port id="5" precision="FP32" names="963">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -53798,15 +15782,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="3523" name="Constant_8371" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="1052" name="Constant_3915" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64" names="2418">
+				<port id="0" precision="I64" names="964">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3524" name="__module.model.gpt_neox.layers.14.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
+		<layer id="1053" name="__module.model.gpt_neox.layers.5.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -53819,7 +15803,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2419,key.85">
+				<port id="2" precision="FP32" names="965">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -53827,8 +15811,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="3525" name="__module.model.gpt_neox.layers.14.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="1054" name="__module.model.gpt_neox.layers.5.attention/aten::cat/Concat_5" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -53836,82 +15820,108 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="present.5.value">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3526" name="__module.model.gpt_neox.layers.14.attention/aten::size/Constant_2" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1055" name="Constant_4297" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3527" name="__module.model.gpt_neox.layers.14.attention/aten::size/Gather_2" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="1056" name="ShapeOf_39067" type="ShapeOf" version="opset3">
+			<data output_type="i64" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="2427" />
+				<port id="1" precision="I64">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1057" name="Constant_39068" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116168" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3528" name="Constant_337793_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
+		<layer id="1058" name="Constant_39069" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="3529" name="Constant_337793" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="1059" name="Gather_39070" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="I64">
+					<dim>4</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="1" precision="FP32" />
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3530" name="Constant_337786" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
-			<output>
+		<layer id="1060" name="Add_39073" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="3531" name="Constant_337788" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
+				<port id="2" precision="I64" names="1012">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3532" name="Constant_337790" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
+		<layer id="1061" name="Constant_4300" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3533" name="Concat_337791" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="1062" name="__module.model.gpt_neox.layers.5.attention/aten::slice/Slice_14" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
 					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -53922,42 +15932,49 @@
 				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="5" precision="FP32" names="1016">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3534" name="Broadcast_337794" type="Broadcast" version="opset3">
-			<data mode="numpy" />
+		<layer id="1063" name="__module.model.gpt_neox.layers.5.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+			<data causal="false" />
 			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
+				<port id="1" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>0</dim>
+					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="3535" name="ReadValue_330449" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.14.keypresent.14.key" variable_type="f32" variable_shape="?,32,?,80" />
-			<input>
-				<port id="0" precision="FP32">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>0</dim>
+					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
+				<port id="3" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="past_key_values.14.key">
+				<port id="4" precision="FP32" names="1017,attn_output.31">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -53965,14 +15982,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="3536" name="Constant_323469" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
+		<layer id="1064" name="__module.model.gpt_neox.layers.5.attention/aten::transpose/Constant" type="Const" version="opset1">
+			<data element_type="i32" shape="4" offset="201116200" size="16" />
 			<output>
-				<port id="0" precision="I64" />
+				<port id="0" precision="I32">
+					<dim>4</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3537" name="Gather_323470" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="1065" name="__module.model.gpt_neox.layers.5.attention/aten::transpose/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -53981,1243 +15999,1083 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I32">
-					<dim>-1</dim>
+					<dim>4</dim>
 				</port>
-				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="3" precision="FP32">
+				<port id="2" precision="FP32" names="1018">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3538" name="__module.model.gpt_neox.layers.14.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="1066" name="Constant_40989" type="Const" version="opset1">
+			<data element_type="i64" shape="3" offset="201116216" size="24" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>3</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1067" name="__module.model.gpt_neox.layers.5.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
+			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>80</dim>
 				</port>
+				<port id="1" precision="I64">
+					<dim>3</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="1021">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3539" name="__module.model.gpt_neox.layers.14.attention/aten::size/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1068" name="self.model.gpt_neox.layers.5.attention.dense.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 2560" offset="595215440" size="6553600" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3540" name="__module.model.gpt_neox.layers.14.attention/aten::size/Gather_3" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="1069" name="Convert_75231" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1070" name="self.model.gpt_neox.layers.5.attention.dense.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="601769040" size="2560" />
+			<output>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1071" name="Convert_75234" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="2429" />
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3541" name="__module.model.gpt_neox.layers.14.attention/aten::add_/Add" type="Add" version="opset1">
+		<layer id="1072" name="self.model.gpt_neox.layers.5.attention.dense.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64" names="2428,2432,2433,seq_len.57" />
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3542" name="__module.model.gpt_neox.layers.14.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1073" name="self.model.gpt_neox.layers.5.attention.dense.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="601771600" size="5120" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3543" name="__module.model.gpt_neox.layers.14.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1">
+		<layer id="1074" name="self.model.gpt_neox.layers.5.attention.dense.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3544" name="__module.model.gpt_neox.layers.14.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1075" name="self.model.gpt_neox.layers.5.attention.dense.weight/fq_weights_1/convert" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3545" name="__module.model.gpt_neox.layers.14.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1076" name="__module.model.gpt_neox.layers.5.attention.dense/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="2" precision="I64">
+			</input>
+			<output>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1077" name="Constant_37335" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="601776720" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
 					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
+			</output>
+		</layer>
+		<layer id="1078" name="__module.model.gpt_neox.layers.5.attention.dense/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
 					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2436,2439,cos.57">
+				<port id="2" precision="FP32" names="1024,1026,input.21">
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3546" name="Constant_38281" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
+		<layer id="1079" name="__module.model.gpt_neox.layers.5/aten::add/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I32" />
+				<port id="2" precision="FP32" names="1044">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3547" name="__module.model.gpt_neox.layers.14.attention/aten::index/Gather" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="1080" name="__module.model.gpt_neox.layers.5/aten::add/Add_1" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I32">
+				<port id="1" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="2442">
+				<port id="2" precision="FP32" names="1045,1047,hidden_states.27">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1081" name="__module.model.gpt_neox.layers.6.input_layernorm/aten::layer_norm/Multiply" type="Const" version="opset1">
+			<data element_type="i32" shape="1" offset="128867668" size="4" />
+			<output>
+				<port id="0" precision="I32">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3548" name="__module.model.gpt_neox.layers.14.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1">
+		<layer id="1082" name="__module.model.gpt_neox.layers.6.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
+			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="I32">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2443,cos.59">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1083" name="Constant_37336" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="601786960" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3549" name="__module.model.gpt_neox.layers.14.attention/aten::mul/Multiply" type="Multiply" version="opset1">
+		<layer id="1084" name="__module.model.gpt_neox.layers.6.post_attention_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2447">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3550" name="Constant_274314" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3551" name="Constant_274313" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1085" name="Constant_37337" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="601797200" size="10240" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="FP32">
+					<dim>1</dim>
 					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3552" name="__module.model.gpt_neox.layers.14.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="1086" name="__module.model.gpt_neox.layers.6.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="1145">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3553" name="__module.model.gpt_neox.layers.14.attention/aten::size/Constant_5" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1087" name="self.model.gpt_neox.layers.6.mlp.dense_h_to_4h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 2560" offset="601807440" size="26214400" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="3554" name="__module.model.gpt_neox.layers.14.attention/aten::size/Gather_5" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="2448,2453" />
-			</output>
-		</layer>
-		<layer id="3555" name="__module.model.gpt_neox.layers.14.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" />
 			</output>
 		</layer>
-		<layer id="3556" name="__module.model.gpt_neox.layers.14.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1">
+		<layer id="1088" name="Convert_74304" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64" />
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64" names="2450,2451,2455,2456" />
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3557" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Constant_1" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1089" name="self.model.gpt_neox.layers.6.mlp.dense_h_to_4h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 1" offset="628021840" size="10240" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3558" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1">
+		<layer id="1090" name="Convert_74307" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="U8">
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="3559" name="Constant_274312" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I32">
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3560" name="ScatterUpdate_274315" type="ScatterUpdate" version="opset3">
+		<layer id="1091" name="self.model.gpt_neox.layers.6.mlp.dense_h_to_4h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3561" name="Constant_274318" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3562" name="Constant_274321" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1092" name="self.model.gpt_neox.layers.6.mlp.dense_h_to_4h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="10240, 1" offset="628032080" size="20480" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3563" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1093" name="self.model.gpt_neox.layers.6.mlp.dense_h_to_4h.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2457,x2.57">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3564" name="Constant_299216_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3565" name="Constant_299216" type="Convert" version="opset1">
+		<layer id="1094" name="__module.model.gpt_neox.layers.6.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
 			</rt_info>
 			<input>
 				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3566" name="__module.model.gpt_neox.layers.14.attention/aten::neg/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1095" name="__module.model.gpt_neox.layers.6.mlp.dense_h_to_4h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2458">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3567" name="Constant_274366" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1096" name="self.model.gpt_neox.layers.6.mlp.dense_h_to_4h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="10240" offset="628052560" size="20480" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.6.mlp.dense_h_to_4h.bias">
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3568" name="Constant_274371" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
+		<layer id="1097" name="__module.model.gpt_neox.layers.6.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<rt_info>
-				<attribute name="precise" version="0" />
+				<attribute name="decompression" version="0" />
 			</rt_info>
+			<input>
+				<port id="0" precision="BF16">
+					<dim>10240</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3569" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1098" name="__module.model.gpt_neox.layers.6.mlp.dense_h_to_4h/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2452">
+				<port id="2" precision="FP32" names="1150">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3570" name="__module.model.gpt_neox.layers.14.attention/aten::cat/Concat" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="1099" name="__module.model.gpt_neox.layers.6.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
+			<data approximation_mode="ERF" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2460">
+				<port id="1" precision="FP32" names="1151">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3571" name="self.model.gpt_neox.layers.14.attention.rotary_emb.sin_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201162356" size="81920" />
+		<layer id="1100" name="self.model.gpt_neox.layers.6.mlp.dense_4h_to_h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 10240" offset="628073040" size="26214400" />
 			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.14.attention.rotary_emb.sin_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3572" name="self.model.gpt_neox.layers.14.attention.rotary_emb.sin_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="1101" name="Convert_74314" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3573" name="__module.model.gpt_neox.layers.14.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1102" name="self.model.gpt_neox.layers.6.mlp.dense_4h_to_h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="654287440" size="2560" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="U8">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3574" name="__module.model.gpt_neox.layers.14.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1103" name="Convert_74317" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3575" name="__module.model.gpt_neox.layers.14.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1104" name="self.model.gpt_neox.layers.6.mlp.dense_4h_to_h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="3" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2437,2440,sin.57">
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3576" name="Constant_38284" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
+		<layer id="1105" name="self.model.gpt_neox.layers.6.mlp.dense_4h_to_h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="654290000" size="5120" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3577" name="__module.model.gpt_neox.layers.14.attention/aten::index/Gather_1" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="1106" name="self.model.gpt_neox.layers.6.mlp.dense_4h_to_h.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
 				</port>
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="2445">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3578" name="__module.model.gpt_neox.layers.14.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1">
+		<layer id="1107" name="__module.model.gpt_neox.layers.6.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<rt_info>
+				<attribute name="decompression" version="0" />
+			</rt_info>
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="1" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2446,sin.59">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3579" name="__module.model.gpt_neox.layers.14.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="2461">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3580" name="__module.model.gpt_neox.layers.14.attention/aten::add/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1108" name="__module.model.gpt_neox.layers.6.mlp.dense_4h_to_h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>10240</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2462">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3581" name="Constant_274417" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3582" name="Constant_274420" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1109" name="self.model.gpt_neox.layers.6.mlp.dense_4h_to_h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="2560" offset="654295120" size="5120" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.6.mlp.dense_4h_to_h.bias">
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3583" name="Constant_274423" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
+		<layer id="1110" name="__module.model.gpt_neox.layers.6.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<rt_info>
-				<attribute name="precise" version="0" />
+				<attribute name="decompression" version="0" />
 			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3584" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="BF16">
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2424">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3585" name="__module.model.gpt_neox.layers.14.attention/aten::cat/Concat_1" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="1111" name="__module.model.gpt_neox.layers.6.mlp.dense_4h_to_h/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2480,2490,query.87,query.89">
+				<port id="2" precision="FP32" names="1154,input.27">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3586" name="Constant_274429" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3587" name="Constant_274432" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3588" name="Constant_274435" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1112" name="Constant_37338" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="654300240" size="10240" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3589" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="2425,k.29">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3590" name="__module.model.gpt_neox.layers.14.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
+		<layer id="1113" name="__module.model.gpt_neox.layers.6.input_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2463">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3591" name="Constant_274439" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3592" name="Constant_274438" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1114" name="Constant_37339" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="654310480" size="10240" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="FP32">
+					<dim>1</dim>
 					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3593" name="__module.model.gpt_neox.layers.14.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="1115" name="__module.model.gpt_neox.layers.6.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="1057,hidden_states.25">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3594" name="__module.model.gpt_neox.layers.14.attention/aten::size/Constant_7" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1116" name="self.model.gpt_neox.layers.6.attention.query_key_value.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 2560" offset="654320720" size="19660800" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3595" name="__module.model.gpt_neox.layers.14.attention/aten::size/Gather_7" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="1117" name="Convert_74890" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="2464,2469" />
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3596" name="__module.model.gpt_neox.layers.14.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
+		<layer id="1118" name="self.model.gpt_neox.layers.6.attention.query_key_value.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 1" offset="673981520" size="7680" />
 			<output>
-				<port id="2" precision="I64" />
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3597" name="__module.model.gpt_neox.layers.14.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1">
+		<layer id="1119" name="Convert_74893" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64" />
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64" names="2466,2467,2471,2472" />
-			</output>
-		</layer>
-		<layer id="3598" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3599" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1">
+		<layer id="1120" name="self.model.gpt_neox.layers.6.attention.query_key_value.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="2" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3600" name="Constant_274437" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1121" name="self.model.gpt_neox.layers.6.attention.query_key_value.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="7680, 1" offset="673989200" size="15360" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3601" name="ScatterUpdate_274440" type="ScatterUpdate" version="opset3">
+		<layer id="1122" name="self.model.gpt_neox.layers.6.attention.query_key_value.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3602" name="Constant_274443" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+		<layer id="1123" name="self.model.gpt_neox.layers.6.attention.query_key_value.weight/fq_weights_1/convert" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="3603" name="Constant_274446" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3604" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1124" name="__module.model.gpt_neox.layers.6.attention.query_key_value/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2473,x2.59">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3605" name="Constant_299217_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3606" name="Constant_299217" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
+		<layer id="1125" name="Constant_37340" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 7680" offset="674004560" size="30720" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3607" name="__module.model.gpt_neox.layers.14.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
+		<layer id="1126" name="__module.model.gpt_neox.layers.6.attention.query_key_value/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>7680</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>7680</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2474">
+				<port id="2" precision="FP32" names="1064,qkv.25">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3608" name="Constant_274491" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3609" name="Constant_274496" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1127" name="__module.model.gpt_neox.layers.6.attention/prim::ListConstruct" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201115992" size="32" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3610" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1128" name="__module.model.gpt_neox.layers.6.attention/aten::view/Reshape" type="Reshape" version="opset1">
+			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>7680</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2468">
+				<port id="2" precision="FP32" names="1068,qkv.27">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3611" name="__module.model.gpt_neox.layers.14.attention/aten::cat/Concat_2" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
+		<layer id="1129" name="Constant_4496" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
+			</output>
+		</layer>
+		<layer id="1130" name="Constant_4498" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="1131" name="Constant_4500" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="2" precision="FP32" names="2476">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3612" name="__module.model.gpt_neox.layers.14.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1132" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Slice" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2477">
+				<port id="5" precision="FP32" names="1069">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3613" name="__module.model.gpt_neox.layers.14.attention/aten::add/Add_1" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1133" name="Constant_4509" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
+			<output>
+				<port id="0" precision="I64" names="1070">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1134" name="__module.model.gpt_neox.layers.6.attention/aten::permute/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2478">
+				<port id="2" precision="FP32" names="1071,query.13">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3614" name="Constant_274541" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1135" name="Constant_4550" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3615" name="Constant_274544" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1136" name="Constant_4552" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3616" name="Constant_274547" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1137" name="Constant_4554" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3617" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1138" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Slice_1" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -55226,121 +17084,83 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="2426">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3618" name="__module.model.gpt_neox.layers.14.attention/aten::cat/Concat_3" type="Concat" version="opset1">
-			<data axis="-1" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2482">
+				<port id="5" precision="FP32" names="1078,q.13">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3619" name="__module.model.gpt_neox.layers.14.attention/aten::to/Convert_1" type="Concat" version="opset1">
-			<data axis="-2" />
+		<layer id="1139" name="__module.model.gpt_neox.layers.6.attention/aten::mul/Multiply" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="present.14.key">
+				<port id="2" precision="FP32" names="1084">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3620" name="Constant_337806_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="3621" name="Constant_337806" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="3622" name="Constant_337799" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
+		<layer id="1140" name="Constant_4708" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3623" name="Constant_337801" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
+		<layer id="1141" name="Constant_4710" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3624" name="Constant_337803" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
+		<layer id="1142" name="Constant_4712" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3625" name="Concat_337804" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="1143" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Slice_2" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -55351,266 +17171,209 @@
 				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
 				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3626" name="Broadcast_337807" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3627" name="ReadValue_330451" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.14.valuepresent.14.value" variable_type="f32" variable_shape="?,32,?,80" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="past_key_values.14.value">
+				<port id="5" precision="FP32" names="1094,x2.25">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3628" name="Constant_323472" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
+		<layer id="1144" name="Constant_37341" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="0" precision="I64" />
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3629" name="Gather_323473" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="1145" name="__module.model.gpt_neox.layers.6.attention/aten::neg/Multiply" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>10</dim>
 				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="3" precision="FP32">
+				<port id="2" precision="FP32" names="1095">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3630" name="Constant_274553" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1146" name="Constant_4605" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3631" name="Constant_274556" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1147" name="Constant_4649" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3632" name="Constant_274559" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1148" name="Constant_4651" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3633" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1149" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Slice_3" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>240</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2420">
-					<dim>-1</dim>
+				<port id="5" precision="FP32" names="1089">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3634" name="Constant_8385" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="2421">
-					<dim>4</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3635" name="__module.model.gpt_neox.layers.14.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
+		<layer id="1150" name="__module.model.gpt_neox.layers.6.attention/aten::cat/Concat" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2422">
+				<port id="2" precision="FP32" names="1097">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3636" name="__module.model.gpt_neox.layers.14.attention/aten::cat/Concat_5" type="Concat" version="opset1">
-			<data axis="-2" />
+		<layer id="1151" name="__module.model.gpt_neox.layers.6.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="present.14.value">
+				<port id="2" precision="FP32" names="1098">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3637" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Slice_11" type="Const" version="opset1">
-			<data element_type="boolean" shape="1, 1, 2048, 2048" offset="201244276" size="4194304" />
-			<output>
-				<port id="0" precision="BOOL" names="2499">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3638" name="Constant_274563" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3639" name="Constant_274562" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162298" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3640" name="__module.model.gpt_neox.layers.14.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="1152" name="__module.model.gpt_neox.layers.6.attention/aten::add/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="1099">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3641" name="__module.model.gpt_neox.layers.14.attention/aten::size/Constant_8" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1153" name="Constant_4560" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3642" name="__module.model.gpt_neox.layers.14.attention/aten::size/Gather_8" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
+		<layer id="1154" name="Constant_4562" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
+			</output>
+		</layer>
+		<layer id="1155" name="Constant_4564" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="3" precision="I64" names="2494" />
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3643" name="__module.model.gpt_neox.layers.14.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="1156" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Slice_4" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -55618,80 +17381,88 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-			</input>
-			<output>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="3644" name="__module.model.gpt_neox.layers.14.attention/aten::size/Constant_9" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I32" />
+				<port id="5" precision="FP32" names="1079">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3645" name="__module.model.gpt_neox.layers.14.attention/aten::size/Gather_9" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="1157" name="__module.model.gpt_neox.layers.6.attention/aten::cat/Concat_1" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="2492" />
+				<port id="2" precision="FP32" names="1117">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3646" name="__module.model.gpt_neox.layers.14.attention/aten::sub/Subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
+		<layer id="1158" name="Constant_52695" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="2" precision="I64" names="2496,2497" />
+				<port id="0" precision="FP32" />
 			</output>
 		</layer>
-		<layer id="3647" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Constant_4" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1159" name="Constant_52688" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3648" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="1160" name="Constant_52690" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3649" name="Constant_274561" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1161" name="Constant_52692" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3650" name="ScatterUpdate_274564" type="ScatterUpdate" version="opset3">
+		<layer id="1162" name="Concat_52693" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
 				<port id="0" precision="I64">
-					<dim>3</dim>
+					<dim>1</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -55699,146 +17470,112 @@
 				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
 				<port id="4" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3651" name="Constant_274565" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3652" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1">
+		<layer id="1163" name="Broadcast_52696" type="Broadcast" version="opset3">
+			<data mode="numpy" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
+					<dim>4</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3653" name="ScatterUpdate_274566" type="ScatterUpdate" version="opset3">
+		<layer id="1164" name="ReadValue_50295" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.6.keypresent.6.key" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
+				<port id="1" precision="FP32" names="past_key_values.6.key">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3654" name="Constant_274569" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438612" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1165" name="Constant_47814" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="3655" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1166" name="Gather_47815" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>3</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>3</dim>
+				<port id="1" precision="I32">
+					<dim>-1</dim>
 				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="4" precision="BOOL" names="2500">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="3" precision="FP32">
 					<dim>-1</dim>
-					<dim>2048</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3656" name="Constant_274642" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3657" name="Constant_274643" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1167" name="Constant_4514" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3658" name="Constant_274639" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1168" name="Constant_4516" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3659" name="Constant_274638" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1169" name="Constant_4518" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3660" name="ScatterUpdate_274644" type="ScatterUpdate" version="opset3">
+		<layer id="1170" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Slice_5" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -55846,183 +17583,161 @@
 				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="5" precision="FP32" names="1072">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3661" name="Constant_274647" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1171" name="Constant_4527" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="I64" names="1073">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3662" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1172" name="__module.model.gpt_neox.layers.6.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
 			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>2048</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
 			</input>
 			<output>
-				<port id="4" precision="BOOL" names="2501,2502">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="1074,key.25">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3663" name="__module.model.gpt_neox.layers.14.attention/aten::where/Convert_1_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
+		<layer id="1173" name="Constant_4570" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3664" name="__module.model.gpt_neox.layers.14.attention/aten::where/Convert_1" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
+		<layer id="1174" name="Constant_4572" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="1" precision="FP32" />
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3665" name="__module.model.gpt_neox.layers.14.attention/aten::full/Convert" type="Const" version="opset1">
-			<data element_type="f32" shape="" offset="205438636" size="4" />
+		<layer id="1175" name="Constant_4574" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="FP32" names="2489" />
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3666" name="__module.model.gpt_neox.layers.14.attention/aten::where/Select" type="Select" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1176" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Slice_6" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32" />
-				<port id="2" precision="FP32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="2503,causal_mask.57">
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="5" precision="FP32" names="1080,k.13">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3667" name="Add_290422" type="Add" version="opset1">
+		<layer id="1177" name="__module.model.gpt_neox.layers.6.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>1</dim>
 					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="1100">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3668" name="ShapeOf_290423" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</input>
+		<layer id="1178" name="Constant_4851" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3669" name="Constant_254539" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1179" name="Constant_4853" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3670" name="__module.model.gpt_neox.layers.14.attention/aten::size/Constant_10" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1180" name="Constant_4855" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="3671" name="__module.model.gpt_neox.layers.14.attention/aten::size/Gather_10" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="2487">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3672" name="__module.model.gpt_neox.layers.14.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="1181" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Slice_7" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -56033,752 +17748,743 @@
 				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64" names="2504">
-					<dim>4</dim>
+				<port id="5" precision="FP32" names="1110,x2.27">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1182" name="Constant_37342" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3673" name="__module.model.gpt_neox.layers.14.attention/aten::expand/Abs" type="Abs" version="opset1">
+		<layer id="1183" name="__module.model.gpt_neox.layers.6.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="1111">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3674" name="Maximum_290424" type="Maximum" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
+		<layer id="1184" name="Constant_4748" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+			</output>
+		</layer>
+		<layer id="1185" name="Constant_4792" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="1186" name="Constant_4794" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3675" name="__module.model.gpt_neox.layers.14.attention/aten::add/Add_2" type="Broadcast" version="opset3">
-			<data mode="numpy" />
+		<layer id="1187" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Slice_8" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2506">
-					<dim>-1</dim>
-					<dim>-1</dim>
+				<port id="5" precision="FP32" names="1105">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3676" name="__module.model.gpt_neox.layers.14.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
-			<data causal="false" />
+		<layer id="1188" name="__module.model.gpt_neox.layers.6.attention/aten::cat/Concat_2" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>10</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2507,2508,sdpa_result.29,tensor.57">
+				<port id="2" precision="FP32" names="1113">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3677" name="Constant_8826" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="2509">
-					<dim>4</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3678" name="__module.model.gpt_neox.layers.14.attention/aten::permute/Transpose_3" type="Transpose" version="opset1">
+		<layer id="1189" name="__module.model.gpt_neox.layers.6.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2510">
-					<dim>-1</dim>
+				<port id="2" precision="FP32" names="1114">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3679" name="__module.model.gpt_neox.layers.14.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438650" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3680" name="__module.model.gpt_neox.layers.14.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
-			<data special_zero="true" />
+		<layer id="1190" name="__module.model.gpt_neox.layers.6.attention/aten::add/Add_1" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>3</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2515">
+				<port id="2" precision="FP32" names="1115">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3681" name="self.model.gpt_neox.layers.14.attention.dense.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 2560" offset="1308342994" size="6553600" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3682" name="Convert_21994" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="1191" name="Constant_4580" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3683" name="self.model.gpt_neox.layers.14.attention.dense.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="1314896594" size="2560" />
+		<layer id="1192" name="Constant_4582" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3684" name="Convert_21997" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
+		<layer id="1193" name="Constant_4584" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3685" name="self.model.gpt_neox.layers.14.attention.dense.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1194" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Slice_9" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3686" name="self.model.gpt_neox.layers.14.attention.dense.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="1314899154" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
+				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="3687" name="self.model.gpt_neox.layers.14.attention.dense.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="3" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="5" precision="FP32" names="1081">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3688" name="self.model.gpt_neox.layers.14.attention.dense.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="1195" name="__module.model.gpt_neox.layers.6.attention/aten::cat/Concat_3" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32" names="1119">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3689" name="__module.model.gpt_neox.layers.14.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
+		<layer id="1196" name="__module.model.gpt_neox.layers.6.attention/aten::cat/Concat_4" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="present.6.key">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3690" name="Constant_299218_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1314904274" size="5120" />
+		<layer id="1197" name="Constant_52707" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="FP32" />
+			</output>
+		</layer>
+		<layer id="1198" name="Constant_52700" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3691" name="Constant_299218" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+		<layer id="1199" name="Constant_52702" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="1200" name="Constant_52704" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3692" name="__module.model.gpt_neox.layers.14.attention.dense/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1201" name="Concat_52705" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="FP32">
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2518,2520,input.59">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
+				<port id="4" precision="I64">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3693" name="__module.model.gpt_neox.layers.14/aten::add/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1202" name="Broadcast_52708" type="Broadcast" version="opset3">
+			<data mode="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2538">
-					<dim>-1</dim>
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3694" name="__module.model.gpt_neox.layers.14/aten::add/Add_1" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1203" name="ReadValue_50297" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.6.valuepresent.6.value" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2539,2541,hidden_states.31">
+				<port id="1" precision="FP32" names="past_key_values.6.value">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3695" name="__module.model.gpt_neox.layers.15.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867670" size="4" />
+		<layer id="1204" name="Constant_47817" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="3696" name="__module.model.gpt_neox.layers.15.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
-			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
+		<layer id="1205" name="Gather_47818" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="I32">
-					<dim>1</dim>
+					<dim>-1</dim>
 				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="3" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3697" name="Constant_299219_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1314909394" size="5120" />
+		<layer id="1206" name="Constant_4532" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3698" name="Constant_299219" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+		<layer id="1207" name="Constant_4534" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="1208" name="Constant_4536" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3699" name="__module.model.gpt_neox.layers.15.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1209" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Slice_10" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
-				<port id="1" precision="FP32">
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="5" precision="FP32" names="1075">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3700" name="Constant_299220_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1314914514" size="5120" />
+		<layer id="1210" name="Constant_4545" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64" names="1076">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3701" name="Constant_299220" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="1211" name="__module.model.gpt_neox.layers.6.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32" names="1077">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3702" name="__module.model.gpt_neox.layers.15.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1212" name="__module.model.gpt_neox.layers.6.attention/aten::cat/Concat_5" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2676">
+				<port id="2" precision="FP32" names="present.6.value">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3703" name="self.model.gpt_neox.layers.15.mlp.dense_h_to_4h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 2560" offset="1314919634" size="26214400" />
+		<layer id="1213" name="Constant_4927" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3704" name="Convert_22044" type="Convert" version="opset1">
-			<data destination_type="f16" />
+		<layer id="1214" name="ShapeOf_39138" type="ShapeOf" version="opset3">
+			<data output_type="i64" />
 			<input>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3705" name="self.model.gpt_neox.layers.15.mlp.dense_h_to_4h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 1" offset="1341134034" size="10240" />
+		<layer id="1215" name="Constant_39139" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116168" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3706" name="Convert_22047" type="Convert" version="opset1">
-			<data destination_type="f16" />
+		<layer id="1216" name="Constant_39140" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64" />
+			</output>
+		</layer>
+		<layer id="1217" name="Gather_39141" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
+				<port id="0" precision="I64">
+					<dim>4</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3707" name="self.model.gpt_neox.layers.15.mlp.dense_h_to_4h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="1218" name="Add_39144" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
+				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="2" precision="I64" names="1124">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3708" name="self.model.gpt_neox.layers.15.mlp.dense_h_to_4h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="10240, 1" offset="1341144274" size="20480" />
+		<layer id="1219" name="Constant_4930" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3709" name="self.model.gpt_neox.layers.15.mlp.dense_h_to_4h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1220" name="__module.model.gpt_neox.layers.6.attention/aten::slice/Slice_14" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
+				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="2" precision="I64">
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="3710" name="self.model.gpt_neox.layers.15.mlp.dense_h_to_4h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="5" precision="FP32" names="1128">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3711" name="__module.model.gpt_neox.layers.15.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
+		<layer id="1221" name="__module.model.gpt_neox.layers.6.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+			<data causal="false" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>10240</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-			</input>
-			<output>
 				<port id="2" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>80</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="3712" name="Constant_299221_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 10240" offset="1341164754" size="20480" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="3" precision="FP32">
+					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>10240</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="4" precision="FP32" names="1129,attn_output.37">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3713" name="Constant_299221" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
+		<layer id="1222" name="__module.model.gpt_neox.layers.6.attention/aten::transpose/Constant" type="Const" version="opset1">
+			<data element_type="i32" shape="4" offset="201116200" size="16" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
+				<port id="0" precision="I32">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3714" name="__module.model.gpt_neox.layers.15.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1223" name="__module.model.gpt_neox.layers.6.attention/aten::transpose/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
+				<port id="1" precision="I32">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2681">
+				<port id="2" precision="FP32" names="1130">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3715" name="__module.model.gpt_neox.layers.15.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
-			<data approximation_mode="ERF" />
+		<layer id="1224" name="Constant_40990" type="Const" version="opset1">
+			<data element_type="i64" shape="3" offset="201116216" size="24" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>3</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1225" name="__module.model.gpt_neox.layers.6.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
+			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>3</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="2682">
+				<port id="2" precision="FP32" names="1133">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3716" name="self.model.gpt_neox.layers.15.mlp.dense_4h_to_h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 10240" offset="1341185234" size="26214400" />
+		<layer id="1226" name="self.model.gpt_neox.layers.6.attention.dense.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 2560" offset="674035280" size="6553600" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>2560</dim>
-					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3717" name="Convert_22054" type="Convert" version="opset1">
+		<layer id="1227" name="Convert_75242" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
 					<dim>2560</dim>
-					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP16">
 					<dim>2560</dim>
-					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3718" name="self.model.gpt_neox.layers.15.mlp.dense_4h_to_h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="1367399634" size="2560" />
+		<layer id="1228" name="self.model.gpt_neox.layers.6.attention.dense.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="680588880" size="2560" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>2560</dim>
@@ -56786,7 +18492,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="3719" name="Convert_22057" type="Convert" version="opset1">
+		<layer id="1229" name="Convert_75245" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -56801,12 +18507,12 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="3720" name="self.model.gpt_neox.layers.15.mlp.dense_4h_to_h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="1230" name="self.model.gpt_neox.layers.6.attention.dense.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
 					<dim>2560</dim>
-					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP16">
 					<dim>2560</dim>
@@ -56816,12 +18522,12 @@
 			<output>
 				<port id="2" precision="FP16">
 					<dim>2560</dim>
-					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3721" name="self.model.gpt_neox.layers.15.mlp.dense_4h_to_h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="1367402194" size="5120" />
+		<layer id="1231" name="self.model.gpt_neox.layers.6.attention.dense.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="680591440" size="5120" />
 			<output>
 				<port id="0" precision="FP16">
 					<dim>2560</dim>
@@ -56829,12 +18535,12 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="3722" name="self.model.gpt_neox.layers.15.mlp.dense_4h_to_h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="1232" name="self.model.gpt_neox.layers.6.attention.dense.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
 					<dim>2560</dim>
-					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP16">
 					<dim>2560</dim>
@@ -56844,39 +18550,36 @@
 			<output>
 				<port id="2" precision="FP16">
 					<dim>2560</dim>
-					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3723" name="self.model.gpt_neox.layers.15.mlp.dense_4h_to_h.weight" type="Convert" version="opset1">
+		<layer id="1233" name="self.model.gpt_neox.layers.6.attention.dense.weight/fq_weights_1/convert" type="Convert" version="opset1">
 			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
 			<input>
 				<port id="0" precision="FP16">
 					<dim>2560</dim>
-					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP32">
 					<dim>2560</dim>
-					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3724" name="__module.model.gpt_neox.layers.15.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="1234" name="__module.model.gpt_neox.layers.6.attention.dense/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>2560</dim>
-					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
@@ -56887,37 +18590,39 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="3725" name="Constant_299222_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1367407314" size="5120" />
+		<layer id="1235" name="Constant_37343" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="680596560" size="10240" />
 			<output>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3726" name="Constant_299222" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="1236" name="__module.model.gpt_neox.layers.6.attention.dense/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="1136,1138,input.25">
+					<dim>-1</dim>
+					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3727" name="__module.model.gpt_neox.layers.15.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1">
+		<layer id="1237" name="__module.model.gpt_neox.layers.6/aten::add/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -56926,61 +18631,59 @@
 					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2685,input.65">
+				<port id="2" precision="FP32" names="1156">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3728" name="Constant_299223_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1367412434" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
+		<layer id="1238" name="__module.model.gpt_neox.layers.6/aten::add/Add_1" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="3729" name="Constant_299223" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="1157,1159,hidden_states.31">
+					<dim>-1</dim>
+					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3730" name="__module.model.gpt_neox.layers.15.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1239" name="__module.model.gpt_neox.layers.7.input_layernorm/aten::layer_norm/Multiply" type="Const" version="opset1">
+			<data element_type="i32" shape="1" offset="128867668" size="4" />
+			<output>
+				<port id="0" precision="I32">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1240" name="__module.model.gpt_neox.layers.7.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
+			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="1" precision="I32">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
@@ -56991,37 +18694,49 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="3731" name="Constant_299224_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1367417554" size="5120" />
+		<layer id="1241" name="Constant_37344" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="680606800" size="10240" />
 			<output>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3732" name="Constant_299224" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="1242" name="__module.model.gpt_neox.layers.7.post_attention_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1243" name="Constant_37345" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="680617040" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3733" name="__module.model.gpt_neox.layers.15.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+		<layer id="1244" name="__module.model.gpt_neox.layers.7.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -57036,127 +18751,127 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2551">
+				<port id="2" precision="FP32" names="1257">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3734" name="self.model.gpt_neox.layers.15.attention.query_key_value.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 2560" offset="1367422674" size="19660800" />
+		<layer id="1245" name="self.model.gpt_neox.layers.7.mlp.dense_h_to_4h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 2560" offset="680627280" size="26214400" />
 			<output>
 				<port id="0" precision="U8">
-					<dim>7680</dim>
+					<dim>10240</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3735" name="Convert_22024" type="Convert" version="opset1">
+		<layer id="1246" name="Convert_74324" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
-					<dim>7680</dim>
+					<dim>10240</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP16">
-					<dim>7680</dim>
+					<dim>10240</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3736" name="self.model.gpt_neox.layers.15.attention.query_key_value.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 1" offset="1387083474" size="7680" />
+		<layer id="1247" name="self.model.gpt_neox.layers.7.mlp.dense_h_to_4h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 1" offset="706841680" size="10240" />
 			<output>
 				<port id="0" precision="U8">
-					<dim>7680</dim>
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3737" name="Convert_22027" type="Convert" version="opset1">
+		<layer id="1248" name="Convert_74327" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
-					<dim>7680</dim>
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP16">
-					<dim>7680</dim>
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3738" name="self.model.gpt_neox.layers.15.attention.query_key_value.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="1249" name="self.model.gpt_neox.layers.7.mlp.dense_h_to_4h.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
-					<dim>7680</dim>
+					<dim>10240</dim>
 					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP16">
-					<dim>7680</dim>
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
 				<port id="2" precision="FP16">
-					<dim>7680</dim>
+					<dim>10240</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3739" name="self.model.gpt_neox.layers.15.attention.query_key_value.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="7680, 1" offset="1387091154" size="15360" />
+		<layer id="1250" name="self.model.gpt_neox.layers.7.mlp.dense_h_to_4h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="10240, 1" offset="706851920" size="20480" />
 			<output>
 				<port id="0" precision="FP16">
-					<dim>7680</dim>
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3740" name="self.model.gpt_neox.layers.15.attention.query_key_value.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="1251" name="self.model.gpt_neox.layers.7.mlp.dense_h_to_4h.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
-					<dim>7680</dim>
+					<dim>10240</dim>
 					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP16">
-					<dim>7680</dim>
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
 				<port id="2" precision="FP16">
-					<dim>7680</dim>
+					<dim>10240</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3741" name="self.model.gpt_neox.layers.15.attention.query_key_value.weight" type="Convert" version="opset1">
+		<layer id="1252" name="__module.model.gpt_neox.layers.7.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
 			</rt_info>
 			<input>
 				<port id="0" precision="FP16">
-					<dim>7680</dim>
+					<dim>10240</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP32">
-					<dim>7680</dim>
+					<dim>10240</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3742" name="__module.model.gpt_neox.layers.15.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="1253" name="__module.model.gpt_neox.layers.7.mlp.dense_h_to_4h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -57165,7 +18880,7 @@
 					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>7680</dim>
+					<dim>10240</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
@@ -57173,826 +18888,536 @@
 				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3743" name="Constant_299225_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 7680" offset="1387106514" size="15360" />
+		<layer id="1254" name="self.model.gpt_neox.layers.7.mlp.dense_h_to_4h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="10240" offset="706872400" size="20480" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.7.mlp.dense_h_to_4h.bias">
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3744" name="Constant_299225" type="Convert" version="opset1">
+		<layer id="1255" name="__module.model.gpt_neox.layers.7.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
 			</rt_info>
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
+				<port id="0" precision="BF16">
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3745" name="__module.model.gpt_neox.layers.15.attention.query_key_value/aten::linear/Add" type="Add" version="opset1">
+		<layer id="1256" name="__module.model.gpt_neox.layers.7.mlp.dense_h_to_4h/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>10240</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2558,qkv.61">
+				<port id="2" precision="FP32" names="1262">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3746" name="__module.model.gpt_neox.layers.15.attention/prim::ListConstruct/Concat" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080154" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3747" name="__module.model.gpt_neox.layers.15.attention/aten::view/Reshape" type="Reshape" version="opset1">
-			<data special_zero="true" />
+		<layer id="1257" name="__module.model.gpt_neox.layers.7.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
+			<data approximation_mode="ERF" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2562,qkv.63">
+				<port id="1" precision="FP32" names="1263">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3748" name="Constant_274654" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3749" name="Constant_274657" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3750" name="Constant_274660" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1258" name="self.model.gpt_neox.layers.7.mlp.dense_4h_to_h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 10240" offset="706892880" size="26214400" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3751" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1259" name="Convert_74334" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2563">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3752" name="Constant_8938" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="1260" name="self.model.gpt_neox.layers.7.mlp.dense_4h_to_h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="733107280" size="2560" />
 			<output>
-				<port id="0" precision="I64" names="2564">
-					<dim>4</dim>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3753" name="__module.model.gpt_neox.layers.15.attention/aten::permute/Transpose" type="Transpose" version="opset1">
+		<layer id="1261" name="Convert_74337" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2565,query.91">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3754" name="Constant_274666" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+		<layer id="1262" name="self.model.gpt_neox.layers.7.mlp.dense_4h_to_h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="3755" name="Constant_274669" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3756" name="Constant_274672" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1263" name="self.model.gpt_neox.layers.7.mlp.dense_4h_to_h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="733109840" size="5120" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3757" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1264" name="self.model.gpt_neox.layers.7.mlp.dense_4h_to_h.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2572,q.31">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3758" name="self.model.gpt_neox.layers.15.attention.rotary_emb.cos_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201080346" size="81920" />
-			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.15.attention.rotary_emb.cos_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3759" name="self.model.gpt_neox.layers.15.attention.rotary_emb.cos_cached" type="Convert" version="opset1">
+		<layer id="1265" name="__module.model.gpt_neox.layers.7.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
 			</rt_info>
 			<input>
 				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3760" name="__module.model.gpt_neox.layers.15.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3761" name="Constant_274678" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3762" name="Constant_274681" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3763" name="Constant_274684" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3764" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1266" name="__module.model.gpt_neox.layers.7.mlp.dense_4h_to_h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2566">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3765" name="Constant_8952" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="1267" name="self.model.gpt_neox.layers.7.mlp.dense_4h_to_h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="2560" offset="733114960" size="5120" />
 			<output>
-				<port id="0" precision="I64" names="2567">
-					<dim>4</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.7.mlp.dense_4h_to_h.bias">
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3766" name="__module.model.gpt_neox.layers.15.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
+		<layer id="1268" name="__module.model.gpt_neox.layers.7.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<rt_info>
+				<attribute name="decompression" version="0" />
+			</rt_info>
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="BF16">
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2568,key.91">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3767" name="__module.model.gpt_neox.layers.15.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="1269" name="__module.model.gpt_neox.layers.7.mlp.dense_4h_to_h/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="3768" name="__module.model.gpt_neox.layers.15.attention/aten::size/Constant_2" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="3769" name="__module.model.gpt_neox.layers.15.attention/aten::size/Gather_2" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="2576" />
-			</output>
-		</layer>
-		<layer id="3770" name="Constant_337819_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="3771" name="Constant_337819" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
 			</input>
 			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="3772" name="Constant_337812" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="1266,input.31">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3773" name="Constant_337814" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
+		<layer id="1270" name="Constant_37346" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="733120080" size="10240" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3774" name="Constant_337816" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
-			<output>
-				<port id="0" precision="I64">
 					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3775" name="Concat_337817" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="1271" name="__module.model.gpt_neox.layers.7.input_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="2" precision="I64">
+				<port id="1" precision="FP32">
 					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
 					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3776" name="Broadcast_337820" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3777" name="ReadValue_330453" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.15.keypresent.15.key" variable_type="f32" variable_shape="?,32,?,80" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32" names="past_key_values.15.key">
-					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3778" name="Constant_323475" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
+		<layer id="1272" name="Constant_37347" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="733130320" size="10240" />
 			<output>
-				<port id="0" precision="I64" />
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3779" name="Gather_323476" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="1273" name="__module.model.gpt_neox.layers.7.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="3" precision="FP32">
+				<port id="2" precision="FP32" names="1169,hidden_states.29">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3780" name="__module.model.gpt_neox.layers.15.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
+		<layer id="1274" name="self.model.gpt_neox.layers.7.attention.query_key_value.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 2560" offset="733140560" size="19660800" />
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3781" name="__module.model.gpt_neox.layers.15.attention/aten::size/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="3782" name="__module.model.gpt_neox.layers.15.attention/aten::size/Gather_3" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="1275" name="Convert_74901" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="2578" />
-			</output>
-		</layer>
-		<layer id="3783" name="__module.model.gpt_neox.layers.15.attention/aten::add_/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="I64" names="2577,2581,2582,seq_len.61" />
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3784" name="__module.model.gpt_neox.layers.15.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1276" name="self.model.gpt_neox.layers.7.attention.query_key_value.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 1" offset="752801360" size="7680" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3785" name="__module.model.gpt_neox.layers.15.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1">
+		<layer id="1277" name="Convert_74904" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="U8">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="3786" name="__module.model.gpt_neox.layers.15.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3787" name="__module.model.gpt_neox.layers.15.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1278" name="self.model.gpt_neox.layers.7.attention.query_key_value.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1279" name="self.model.gpt_neox.layers.7.attention.query_key_value.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="7680, 1" offset="752809040" size="15360" />
+			<output>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I64">
+			</output>
+		</layer>
+		<layer id="1280" name="self.model.gpt_neox.layers.7.attention.query_key_value.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2585,2588,cos.61">
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="2" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3788" name="Constant_38287" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
+		<layer id="1281" name="self.model.gpt_neox.layers.7.attention.query_key_value.weight/fq_weights_1/convert" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I32" />
+				<port id="1" precision="FP32">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3789" name="__module.model.gpt_neox.layers.15.attention/aten::index/Gather" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="1282" name="__module.model.gpt_neox.layers.7.attention.query_key_value/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="2591">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3790" name="__module.model.gpt_neox.layers.15.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64" />
-			</input>
+		<layer id="1283" name="Constant_37348" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 7680" offset="752824400" size="30720" />
 			<output>
-				<port id="2" precision="FP32" names="2592,cos.63">
-					<dim>-1</dim>
+				<port id="0" precision="FP32">
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>1</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3791" name="__module.model.gpt_neox.layers.15.attention/aten::mul/Multiply" type="Multiply" version="opset1">
+		<layer id="1284" name="__module.model.gpt_neox.layers.7.attention.query_key_value/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>7680</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>1</dim>
+					<dim>7680</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2596">
+				<port id="2" precision="FP32" names="1176,qkv.29">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3792" name="Constant_274689" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1285" name="__module.model.gpt_neox.layers.7.attention/prim::ListConstruct" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201115992" size="32" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3793" name="Constant_274688" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3794" name="__module.model.gpt_neox.layers.15.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="1286" name="__module.model.gpt_neox.layers.7.attention/aten::view/Reshape" type="Reshape" version="opset1">
+			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>7680</dim>
 				</port>
-			</input>
-			<output>
 				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="3795" name="__module.model.gpt_neox.layers.15.attention/aten::size/Constant_5" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="3796" name="__module.model.gpt_neox.layers.15.attention/aten::size/Gather_5" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="2597,2602" />
-			</output>
-		</layer>
-		<layer id="3797" name="__module.model.gpt_neox.layers.15.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" />
-			</output>
-		</layer>
-		<layer id="3798" name="__module.model.gpt_neox.layers.15.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
 			</input>
 			<output>
-				<port id="1" precision="I64" names="2599,2600,2604,2605" />
+				<port id="2" precision="FP32" names="1180,qkv.31">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>240</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3799" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Constant_1" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1287" name="Constant_5126" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3800" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="1288" name="Constant_5128" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3801" name="Constant_274687" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1289" name="Constant_5130" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3802" name="ScatterUpdate_274690" type="ScatterUpdate" version="opset3">
+		<layer id="1290" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Slice" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -58000,148 +19425,155 @@
 				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
 				<port id="4" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="3803" name="Constant_274693" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="5" precision="FP32" names="1181">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3804" name="Constant_274696" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1291" name="Constant_5139" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="I64" names="1182">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3805" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1292" name="__module.model.gpt_neox.layers.7.attention/aten::permute/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2606,x2.61">
+				<port id="2" precision="FP32" names="1183,query.15">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3806" name="Constant_299226_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
+		<layer id="1293" name="Constant_5180" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1294" name="Constant_5182" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1295" name="Constant_5184" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3807" name="Constant_299226" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="1296" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Slice_1" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="5" precision="FP32" names="1190,q.15">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3808" name="__module.model.gpt_neox.layers.15.attention/aten::neg/Multiply" type="Multiply" version="opset1">
+		<layer id="1297" name="__module.model.gpt_neox.layers.7.attention/aten::mul/Multiply" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
+					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2607">
+				<port id="2" precision="FP32" names="1196">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3809" name="Constant_274741" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1298" name="Constant_5338" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3810" name="Constant_274746" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1299" name="Constant_5340" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3811" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1300" name="Constant_5342" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1301" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Slice_2" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -58150,17 +19582,20 @@
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2601">
+				<port id="5" precision="FP32" names="1206,x2.29">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -58168,8 +19603,19 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="3812" name="__module.model.gpt_neox.layers.15.attention/aten::cat/Concat" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="1302" name="Constant_37349" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1303" name="__module.model.gpt_neox.layers.7.attention/aten::neg/Multiply" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -58178,75 +19624,51 @@
 					<dim>10</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2609">
+				<port id="2" precision="FP32" names="1207">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3813" name="self.model.gpt_neox.layers.15.attention.rotary_emb.sin_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201162356" size="81920" />
-			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.15.attention.rotary_emb.sin_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3814" name="self.model.gpt_neox.layers.15.attention.rotary_emb.sin_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</input>
+		<layer id="1304" name="Constant_5235" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3815" name="__module.model.gpt_neox.layers.15.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1305" name="Constant_5279" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3816" name="__module.model.gpt_neox.layers.15.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1306" name="Constant_5281" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3817" name="__module.model.gpt_neox.layers.15.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1307" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Slice_3" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>2048</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
@@ -58258,60 +19680,45 @@
 				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2586,2589,sin.61">
+				<port id="5" precision="FP32" names="1201">
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3818" name="Constant_38290" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="3819" name="__module.model.gpt_neox.layers.15.attention/aten::index/Gather_1" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="1308" name="__module.model.gpt_neox.layers.7.attention/aten::cat/Concat" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="2594">
-					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>10</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="3820" name="__module.model.gpt_neox.layers.15.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
+				<port id="1" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>10</dim>
 				</port>
-				<port id="1" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2595,sin.63">
+				<port id="2" precision="FP32" names="1209">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3821" name="__module.model.gpt_neox.layers.15.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
+		<layer id="1309" name="__module.model.gpt_neox.layers.7.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -58328,7 +19735,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2610">
+				<port id="2" precision="FP32" names="1210">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -58336,7 +19743,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="3822" name="__module.model.gpt_neox.layers.15.attention/aten::add/Add" type="Add" version="opset1">
+		<layer id="1310" name="__module.model.gpt_neox.layers.7.attention/aten::add/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -58353,7 +19760,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2611">
+				<port id="2" precision="FP32" names="1211">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -58361,41 +19768,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="3823" name="Constant_274792" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1311" name="Constant_5190" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3824" name="Constant_274795" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1312" name="Constant_5192" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3825" name="Constant_274798" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1313" name="Constant_5194" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3826" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1314" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Slice_4" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -58404,17 +19801,20 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2573">
+				<port id="5" precision="FP32" names="1191">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -58422,7 +19822,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="3827" name="__module.model.gpt_neox.layers.15.attention/aten::cat/Concat_1" type="Concat" version="opset1">
+		<layer id="1315" name="__module.model.gpt_neox.layers.7.attention/aten::cat/Concat_1" type="Concat" version="opset1">
 			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
@@ -58439,7 +19839,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2629,2639,query.93,query.95">
+				<port id="2" precision="FP32" names="1229">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -58447,205 +19847,237 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="3828" name="Constant_274804" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1316" name="Constant_52719" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
+			<output>
+				<port id="0" precision="FP32" />
+			</output>
+		</layer>
+		<layer id="1317" name="Constant_52712" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3829" name="Constant_274807" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1318" name="Constant_52714" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3830" name="Constant_274810" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1319" name="Constant_52716" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3831" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1320" name="Concat_52717" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="4" precision="I64">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1321" name="Broadcast_52720" type="Broadcast" version="opset3">
+			<data mode="numpy" />
+			<input>
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2574,k.31">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3832" name="__module.model.gpt_neox.layers.15.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1322" name="ReadValue_50299" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.7.keypresent.7.key" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP32" names="past_key_values.7.key">
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
+			</output>
+		</layer>
+		<layer id="1323" name="Constant_47820" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64" />
+			</output>
+		</layer>
+		<layer id="1324" name="Gather_47821" type="Gather" version="opset8">
+			<data batch_dims="0" />
+			<input>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
+				<port id="1" precision="I32">
+					<dim>-1</dim>
+				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2612">
+				<port id="3" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3833" name="Constant_274814" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1325" name="Constant_5144" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3834" name="Constant_274813" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1326" name="Constant_5146" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3835" name="__module.model.gpt_neox.layers.15.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="1327" name="Constant_5148" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1328" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Slice_5" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
-			</input>
-			<output>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="3836" name="__module.model.gpt_neox.layers.15.attention/aten::size/Constant_7" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="3837" name="__module.model.gpt_neox.layers.15.attention/aten::size/Gather_7" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="2613,2618" />
+				<port id="5" precision="FP32" names="1184">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3838" name="__module.model.gpt_neox.layers.15.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
+		<layer id="1329" name="Constant_5157" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="2" precision="I64" />
+				<port id="0" precision="I64" names="1185">
+					<dim>4</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3839" name="__module.model.gpt_neox.layers.15.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1">
+		<layer id="1330" name="__module.model.gpt_neox.layers.7.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
 			<input>
-				<port id="0" precision="I64" />
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>4</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64" names="2615,2616,2620,2621" />
+				<port id="2" precision="FP32" names="1186,key.29">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3840" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1331" name="Constant_5200" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3841" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="1332" name="Constant_5202" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3842" name="Constant_274812" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1333" name="Constant_5204" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3843" name="ScatterUpdate_274815" type="ScatterUpdate" version="opset3">
+		<layer id="1334" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Slice_6" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -58653,40 +20085,72 @@
 				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="5" precision="FP32" names="1192,k.15">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3844" name="Constant_274818" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1335" name="__module.model.gpt_neox.layers.7.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="1212">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1336" name="Constant_5481" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3845" name="Constant_274821" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1337" name="Constant_5483" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1338" name="Constant_5485" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3846" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1339" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Slice_7" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -58695,17 +20159,20 @@
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2622,x2.63">
+				<port id="5" precision="FP32" names="1222,x2.31">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -58713,32 +20180,10 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="3847" name="Constant_299227_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3848" name="Constant_299227" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
+		<layer id="1340" name="Constant_37350" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>1</dim>
@@ -58746,7 +20191,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="3849" name="__module.model.gpt_neox.layers.15.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
+		<layer id="1341" name="__module.model.gpt_neox.layers.7.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -58763,7 +20208,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2623">
+				<port id="2" precision="FP32" names="1223">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -58771,30 +20216,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="3850" name="Constant_274866" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1342" name="Constant_5378" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3851" name="Constant_274871" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1343" name="Constant_5422" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1344" name="Constant_5424" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3852" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1345" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Slice_8" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -58803,17 +20249,20 @@
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2617">
+				<port id="5" precision="FP32" names="1217">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -58821,7 +20270,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="3853" name="__module.model.gpt_neox.layers.15.attention/aten::cat/Concat_2" type="Concat" version="opset1">
+		<layer id="1346" name="__module.model.gpt_neox.layers.7.attention/aten::cat/Concat_2" type="Concat" version="opset1">
 			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
@@ -58838,7 +20287,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2625">
+				<port id="2" precision="FP32" names="1225">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -58846,7 +20295,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="3854" name="__module.model.gpt_neox.layers.15.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
+		<layer id="1347" name="__module.model.gpt_neox.layers.7.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -58863,7 +20312,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2626">
+				<port id="2" precision="FP32" names="1226">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -58871,7 +20320,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="3855" name="__module.model.gpt_neox.layers.15.attention/aten::add/Add_1" type="Add" version="opset1">
+		<layer id="1348" name="__module.model.gpt_neox.layers.7.attention/aten::add/Add_1" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -58888,7 +20337,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2627">
+				<port id="2" precision="FP32" names="1227">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -58896,41 +20345,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="3856" name="Constant_274916" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1349" name="Constant_5210" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3857" name="Constant_274919" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1350" name="Constant_5212" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3858" name="Constant_274922" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1351" name="Constant_5214" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3859" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1352" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Slice_9" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -58939,17 +20378,20 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2575">
+				<port id="5" precision="FP32" names="1193">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -58957,7 +20399,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="3860" name="__module.model.gpt_neox.layers.15.attention/aten::cat/Concat_3" type="Concat" version="opset1">
+		<layer id="1353" name="__module.model.gpt_neox.layers.7.attention/aten::cat/Concat_3" type="Concat" version="opset1">
 			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
@@ -58974,7 +20416,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2631">
+				<port id="2" precision="FP32" names="1231">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -58982,7 +20424,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="3861" name="__module.model.gpt_neox.layers.15.attention/aten::to/Convert_1" type="Concat" version="opset1">
+		<layer id="1354" name="__module.model.gpt_neox.layers.7.attention/aten::cat/Concat_4" type="Concat" version="opset1">
 			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
@@ -58999,7 +20441,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="present.15.key">
+				<port id="2" precision="FP32" names="present.7.key">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -59007,49 +20449,37 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="3862" name="Constant_337832_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="3863" name="Constant_337832" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
+		<layer id="1355" name="Constant_52731" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="1" precision="FP32" />
+				<port id="0" precision="FP32" />
 			</output>
 		</layer>
-		<layer id="3864" name="Constant_337825" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
+		<layer id="1356" name="Constant_52724" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3865" name="Constant_337827" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
+		<layer id="1357" name="Constant_52726" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3866" name="Constant_337829" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
+		<layer id="1358" name="Constant_52728" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3867" name="Concat_337830" type="Concat" version="opset1">
+		<layer id="1359" name="Concat_52729" type="Concat" version="opset1">
 			<data axis="0" />
 			<input>
 				<port id="0" precision="I64">
@@ -59071,7 +20501,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="3868" name="Broadcast_337833" type="Broadcast" version="opset3">
+		<layer id="1360" name="Broadcast_52732" type="Broadcast" version="opset3">
 			<data mode="numpy" />
 			<input>
 				<port id="0" precision="FP32" />
@@ -59088,8 +20518,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="3869" name="ReadValue_330455" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.15.valuepresent.15.value" variable_type="f32" variable_shape="?,32,?,80" />
+		<layer id="1361" name="ReadValue_50301" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.7.valuepresent.7.value" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -59099,7 +20529,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="past_key_values.15.value">
+				<port id="1" precision="FP32" names="past_key_values.7.value">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -59107,13 +20537,13 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="3870" name="Constant_323478" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
+		<layer id="1362" name="Constant_47823" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="3871" name="Gather_323479" type="Gather" version="opset8">
+		<layer id="1363" name="Gather_47824" type="Gather" version="opset8">
 			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="FP32">
@@ -59136,885 +20566,1165 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="3872" name="Constant_274928" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1364" name="Constant_5162" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1365" name="Constant_5164" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1366" name="Constant_5166" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1367" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Slice_10" type="Slice" version="opset8">
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>240</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="5" precision="FP32" names="1187">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1368" name="Constant_5175" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
+			<output>
+				<port id="0" precision="I64" names="1188">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3873" name="Constant_274931" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1369" name="__module.model.gpt_neox.layers.7.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>4</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="1189">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1370" name="__module.model.gpt_neox.layers.7.attention/aten::cat/Concat_5" type="Concat" version="opset1">
+			<data axis="-2" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="present.7.value">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1371" name="Constant_5557" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1372" name="ShapeOf_39209" type="ShapeOf" version="opset3">
+			<data output_type="i64" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="I64">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1373" name="Constant_39210" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116168" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1374" name="Constant_39211" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64" />
+			</output>
+		</layer>
+		<layer id="1375" name="Gather_39212" type="Gather" version="opset8">
+			<data batch_dims="0" />
+			<input>
+				<port id="0" precision="I64">
+					<dim>4</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64" />
+			</input>
+			<output>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1376" name="Add_39215" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="I64" names="1236">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1377" name="Constant_5560" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1378" name="__module.model.gpt_neox.layers.7.attention/aten::slice/Slice_14" type="Slice" version="opset8">
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="5" precision="FP32" names="1240">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1379" name="__module.model.gpt_neox.layers.7.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+			<data causal="false" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="3" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="4" precision="FP32" names="1241,attn_output.43">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1380" name="__module.model.gpt_neox.layers.7.attention/aten::transpose/Constant" type="Const" version="opset1">
+			<data element_type="i32" shape="4" offset="201116200" size="16" />
+			<output>
+				<port id="0" precision="I32">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1381" name="__module.model.gpt_neox.layers.7.attention/aten::transpose/Transpose" type="Transpose" version="opset1">
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I32">
 					<dim>4</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="3874" name="Constant_274934" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="1242">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1382" name="Constant_40991" type="Const" version="opset1">
+			<data element_type="i64" shape="3" offset="201116216" size="24" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>3</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1383" name="__module.model.gpt_neox.layers.7.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
+			<data special_zero="true" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>3</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="1245">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1384" name="self.model.gpt_neox.layers.7.attention.dense.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 2560" offset="752855120" size="6553600" />
+			<output>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1385" name="Convert_75253" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1386" name="self.model.gpt_neox.layers.7.attention.dense.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="759408720" size="2560" />
+			<output>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1387" name="Convert_75256" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1388" name="self.model.gpt_neox.layers.7.attention.dense.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1389" name="self.model.gpt_neox.layers.7.attention.dense.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="759411280" size="5120" />
+			<output>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1390" name="self.model.gpt_neox.layers.7.attention.dense.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1391" name="self.model.gpt_neox.layers.7.attention.dense.weight/fq_weights_1/convert" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3875" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1392" name="__module.model.gpt_neox.layers.7.attention.dense/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2569">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3876" name="Constant_8966" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="1393" name="Constant_37351" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="759416400" size="10240" />
 			<output>
-				<port id="0" precision="I64" names="2570">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3877" name="__module.model.gpt_neox.layers.15.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
+		<layer id="1394" name="__module.model.gpt_neox.layers.7.attention.dense/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2571">
+				<port id="2" precision="FP32" names="1248,1250,input.29">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3878" name="__module.model.gpt_neox.layers.15.attention/aten::cat/Concat_5" type="Concat" version="opset1">
-			<data axis="-2" />
+		<layer id="1395" name="__module.model.gpt_neox.layers.7/aten::add/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="present.15.value">
+				<port id="2" precision="FP32" names="1268">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3879" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Slice_11" type="Const" version="opset1">
-			<data element_type="boolean" shape="1, 1, 2048, 2048" offset="201244276" size="4194304" />
-			<output>
-				<port id="0" precision="BOOL" names="2648">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3880" name="Constant_274938" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3881" name="Constant_274937" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162298" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3882" name="__module.model.gpt_neox.layers.15.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="1396" name="__module.model.gpt_neox.layers.7/aten::add/Add_1" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="1269,1271,hidden_states.35">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3883" name="__module.model.gpt_neox.layers.15.attention/aten::size/Constant_8" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1397" name="__module.model.gpt_neox.layers.8.input_layernorm/aten::layer_norm/Multiply" type="Const" version="opset1">
+			<data element_type="i32" shape="1" offset="128867668" size="4" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="3884" name="__module.model.gpt_neox.layers.15.attention/aten::size/Gather_8" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="I32">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="2643" />
 			</output>
 		</layer>
-		<layer id="3885" name="__module.model.gpt_neox.layers.15.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="1398" name="__module.model.gpt_neox.layers.8.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
+			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="I32">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3886" name="__module.model.gpt_neox.layers.15.attention/aten::size/Constant_9" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1399" name="Constant_37352" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="759426640" size="10240" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="3887" name="__module.model.gpt_neox.layers.15.attention/aten::size/Gather_9" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="2641" />
 			</output>
 		</layer>
-		<layer id="3888" name="__module.model.gpt_neox.layers.15.attention/aten::sub/Subtract" type="Subtract" version="opset1">
+		<layer id="1400" name="__module.model.gpt_neox.layers.8.post_attention_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" names="2645,2646" />
-			</output>
-		</layer>
-		<layer id="3889" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Constant_4" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="3890" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3891" name="Constant_274936" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1401" name="Constant_37353" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="759436880" size="10240" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3892" name="ScatterUpdate_274939" type="ScatterUpdate" version="opset3">
+		<layer id="1402" name="__module.model.gpt_neox.layers.8.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="2" precision="I64">
+				<port id="1" precision="FP32">
 					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
 					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3893" name="Constant_274940" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="2" precision="FP32" names="1369">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3894" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="1403" name="self.model.gpt_neox.layers.8.mlp.dense_h_to_4h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 2560" offset="759447120" size="26214400" />
 			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3895" name="ScatterUpdate_274941" type="ScatterUpdate" version="opset3">
+		<layer id="1404" name="Convert_74344" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3896" name="Constant_274944" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438612" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1405" name="self.model.gpt_neox.layers.8.mlp.dense_h_to_4h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 1" offset="785661520" size="10240" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3897" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1406" name="Convert_74347" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
 					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>3</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="BOOL" names="2649">
-					<dim>1</dim>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3898" name="Constant_275017" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+		<layer id="1407" name="self.model.gpt_neox.layers.8.mlp.dense_h_to_4h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="3899" name="Constant_275018" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="3900" name="Constant_275014" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3901" name="Constant_275013" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1408" name="self.model.gpt_neox.layers.8.mlp.dense_h_to_4h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="10240, 1" offset="785671760" size="20480" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3902" name="ScatterUpdate_275019" type="ScatterUpdate" version="opset3">
+		<layer id="1409" name="self.model.gpt_neox.layers.8.mlp.dense_h_to_4h.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3903" name="Constant_275022" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
+		<layer id="1410" name="__module.model.gpt_neox.layers.8.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<rt_info>
-				<attribute name="precise" version="0" />
+				<attribute name="decompression" version="0" />
 			</rt_info>
+			<input>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3904" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1411" name="__module.model.gpt_neox.layers.8.mlp.dense_h_to_4h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="BOOL" names="2650,2651">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3905" name="__module.model.gpt_neox.layers.15.attention/aten::where/Convert_1_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
+		<layer id="1412" name="self.model.gpt_neox.layers.8.mlp.dense_h_to_4h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="10240" offset="785692240" size="20480" />
 			<output>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.8.mlp.dense_h_to_4h.bias">
+					<dim>10240</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3906" name="__module.model.gpt_neox.layers.15.attention/aten::where/Convert_1" type="Convert" version="opset1">
+		<layer id="1413" name="__module.model.gpt_neox.layers.8.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
 			</rt_info>
 			<input>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="BF16">
+					<dim>10240</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="3907" name="__module.model.gpt_neox.layers.15.attention/aten::full/Convert" type="Const" version="opset1">
-			<data element_type="f32" shape="" offset="205438636" size="4" />
-			<output>
-				<port id="0" precision="FP32" names="2638" />
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3908" name="__module.model.gpt_neox.layers.15.attention/aten::where/Select" type="Select" version="opset1">
+		<layer id="1414" name="__module.model.gpt_neox.layers.8.mlp.dense_h_to_4h/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>10240</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
 				</port>
-				<port id="1" precision="FP32" />
-				<port id="2" precision="FP32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="2652,causal_mask.61">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="1374">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3909" name="Add_290428" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1415" name="__module.model.gpt_neox.layers.8.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
+			<data approximation_mode="ERF" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="1" precision="FP32">
 					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
+				<port id="1" precision="FP32" names="1375">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3910" name="ShapeOf_290429" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="1416" name="self.model.gpt_neox.layers.8.mlp.dense_4h_to_h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 10240" offset="785712720" size="26214400" />
+			<output>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1417" name="Convert_74354" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3911" name="Constant_254554" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1418" name="self.model.gpt_neox.layers.8.mlp.dense_4h_to_h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="811927120" size="2560" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="U8">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3912" name="__module.model.gpt_neox.layers.15.attention/aten::size/Constant_10" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="3913" name="__module.model.gpt_neox.layers.15.attention/aten::size/Gather_10" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="1419" name="Convert_74357" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
+				<port id="0" precision="U8">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="2636">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3914" name="__module.model.gpt_neox.layers.15.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="1420" name="self.model.gpt_neox.layers.8.mlp.dense_4h_to_h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1421" name="self.model.gpt_neox.layers.8.mlp.dense_4h_to_h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="811929680" size="5120" />
+			<output>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I64">
+			</output>
+		</layer>
+		<layer id="1422" name="self.model.gpt_neox.layers.8.mlp.dense_4h_to_h.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64" names="2653">
-					<dim>4</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3915" name="__module.model.gpt_neox.layers.15.attention/aten::expand/Abs" type="Abs" version="opset1">
+		<layer id="1423" name="__module.model.gpt_neox.layers.8.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<rt_info>
+				<attribute name="decompression" version="0" />
+			</rt_info>
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3916" name="Maximum_290430" type="Maximum" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1424" name="__module.model.gpt_neox.layers.8.mlp.dense_4h_to_h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1425" name="self.model.gpt_neox.layers.8.mlp.dense_4h_to_h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="2560" offset="811934800" size="5120" />
+			<output>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.8.mlp.dense_4h_to_h.bias">
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3917" name="__module.model.gpt_neox.layers.15.attention/aten::add/Add_2" type="Broadcast" version="opset3">
-			<data mode="numpy" />
+		<layer id="1426" name="__module.model.gpt_neox.layers.8.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<rt_info>
+				<attribute name="decompression" version="0" />
+			</rt_info>
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="BF16">
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2655">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3918" name="__module.model.gpt_neox.layers.15.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
-			<data causal="false" />
+		<layer id="1427" name="__module.model.gpt_neox.layers.8.mlp.dense_4h_to_h/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2656,2657,sdpa_result.31,tensor.61">
+				<port id="2" precision="FP32" names="1378,input.35">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3919" name="Constant_9407" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="1428" name="Constant_37354" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="811939920" size="10240" />
 			<output>
-				<port id="0" precision="I64" names="2658">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3920" name="__module.model.gpt_neox.layers.15.attention/aten::permute/Transpose_3" type="Transpose" version="opset1">
+		<layer id="1429" name="__module.model.gpt_neox.layers.8.input_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2659">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3921" name="__module.model.gpt_neox.layers.15.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438650" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1430" name="Constant_37355" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="811950160" size="10240" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3922" name="__module.model.gpt_neox.layers.15.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
-			<data special_zero="true" />
+		<layer id="1431" name="__module.model.gpt_neox.layers.8.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>3</dim>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2664">
+				<port id="2" precision="FP32" names="1281,hidden_states.33">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3923" name="self.model.gpt_neox.layers.15.attention.dense.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 2560" offset="1387121874" size="6553600" />
+		<layer id="1432" name="self.model.gpt_neox.layers.8.attention.query_key_value.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 2560" offset="811960400" size="19660800" />
 			<output>
 				<port id="0" precision="U8">
-					<dim>2560</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3924" name="Convert_22034" type="Convert" version="opset1">
+		<layer id="1433" name="Convert_74912" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
-					<dim>2560</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP16">
-					<dim>2560</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3925" name="self.model.gpt_neox.layers.15.attention.dense.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="1393675474" size="2560" />
+		<layer id="1434" name="self.model.gpt_neox.layers.8.attention.query_key_value.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 1" offset="831621200" size="7680" />
 			<output>
 				<port id="0" precision="U8">
-					<dim>2560</dim>
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3926" name="Convert_22037" type="Convert" version="opset1">
+		<layer id="1435" name="Convert_74915" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
-					<dim>2560</dim>
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP16">
-					<dim>2560</dim>
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3927" name="self.model.gpt_neox.layers.15.attention.dense.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="1436" name="self.model.gpt_neox.layers.8.attention.query_key_value.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
-					<dim>2560</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP16">
-					<dim>2560</dim>
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
 				<port id="2" precision="FP16">
-					<dim>2560</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3928" name="self.model.gpt_neox.layers.15.attention.dense.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="1393678034" size="5120" />
+		<layer id="1437" name="self.model.gpt_neox.layers.8.attention.query_key_value.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="7680, 1" offset="831628880" size="15360" />
 			<output>
 				<port id="0" precision="FP16">
-					<dim>2560</dim>
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3929" name="self.model.gpt_neox.layers.15.attention.dense.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="1438" name="self.model.gpt_neox.layers.8.attention.query_key_value.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
-					<dim>2560</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP16">
-					<dim>2560</dim>
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
 				<port id="2" precision="FP16">
-					<dim>2560</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3930" name="self.model.gpt_neox.layers.15.attention.dense.weight" type="Convert" version="opset1">
+		<layer id="1439" name="self.model.gpt_neox.layers.8.attention.query_key_value.weight/fq_weights_1/convert" type="Convert" version="opset1">
 			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
 			<input>
 				<port id="0" precision="FP16">
-					<dim>2560</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP32">
-					<dim>2560</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3931" name="__module.model.gpt_neox.layers.15.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="1440" name="__module.model.gpt_neox.layers.8.attention.query_key_value/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -60023,7 +21733,7 @@
 					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>2560</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
@@ -60031,1032 +21741,1214 @@
 				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3932" name="Constant_299228_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1393683154" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3933" name="Constant_299228" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="1441" name="Constant_37356" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 7680" offset="831644240" size="30720" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3934" name="__module.model.gpt_neox.layers.15.attention.dense/aten::linear/Add" type="Add" version="opset1">
+		<layer id="1442" name="__module.model.gpt_neox.layers.8.attention.query_key_value/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>7680</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>7680</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2667,2669,input.63">
+				<port id="2" precision="FP32" names="1288,qkv.33">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3935" name="__module.model.gpt_neox.layers.15/aten::add/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1443" name="__module.model.gpt_neox.layers.8.attention/prim::ListConstruct" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201115992" size="32" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1444" name="__module.model.gpt_neox.layers.8.attention/aten::view/Reshape" type="Reshape" version="opset1">
+			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>7680</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2687">
+				<port id="2" precision="FP32" names="1292,qkv.35">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3936" name="__module.model.gpt_neox.layers.15/aten::add/Add_1" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1445" name="Constant_5756" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1446" name="Constant_5758" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1447" name="Constant_5760" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1448" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Slice" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2688,2690,hidden_states.33">
+				<port id="5" precision="FP32" names="1293">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3937" name="__module.model.gpt_neox.layers.16.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867670" size="4" />
+		<layer id="1449" name="Constant_5769" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
+				<port id="0" precision="I64" names="1294">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3938" name="__module.model.gpt_neox.layers.16.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
-			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
+		<layer id="1450" name="__module.model.gpt_neox.layers.8.attention/aten::permute/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="I32">
-					<dim>1</dim>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="1295,query.17">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3939" name="Constant_299229_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1393688274" size="5120" />
+		<layer id="1451" name="Constant_5810" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="FP16">
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1452" name="Constant_5812" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3940" name="Constant_299229" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="1453" name="Constant_5814" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1454" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Slice_1" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
+				<port id="3" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="5" precision="FP32" names="1302,q.17">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3941" name="__module.model.gpt_neox.layers.16.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
+		<layer id="1455" name="__module.model.gpt_neox.layers.8.attention/aten::mul/Multiply" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
+					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="1308">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3942" name="Constant_299230_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1393693394" size="5120" />
+		<layer id="1456" name="Constant_5968" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
-				<port id="0" precision="FP16">
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1457" name="Constant_5970" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3943" name="Constant_299230" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="1458" name="Constant_5972" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1459" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Slice_2" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
+				<port id="5" precision="FP32" names="1318,x2.33">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1460" name="Constant_37357" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3944" name="__module.model.gpt_neox.layers.16.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+		<layer id="1461" name="__module.model.gpt_neox.layers.8.attention/aten::neg/Multiply" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2825">
+				<port id="2" precision="FP32" names="1319">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3945" name="self.model.gpt_neox.layers.16.mlp.dense_h_to_4h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 2560" offset="1393698514" size="26214400" />
+		<layer id="1462" name="Constant_5865" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3946" name="Convert_22084" type="Convert" version="opset1">
-			<data destination_type="f16" />
+		<layer id="1463" name="Constant_5909" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1464" name="Constant_5911" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1465" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Slice_3" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="5" precision="FP32" names="1313">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3947" name="self.model.gpt_neox.layers.16.mlp.dense_h_to_4h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 1" offset="1419912914" size="10240" />
+		<layer id="1466" name="__module.model.gpt_neox.layers.8.attention/aten::cat/Concat" type="Concat" version="opset1">
+			<data axis="-1" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="1321">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3948" name="Convert_22087" type="Convert" version="opset1">
-			<data destination_type="f16" />
+		<layer id="1467" name="__module.model.gpt_neox.layers.8.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
 					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="1322">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3949" name="self.model.gpt_neox.layers.16.mlp.dense_h_to_4h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="1468" name="__module.model.gpt_neox.layers.8.attention/aten::add/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32" names="1323">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3950" name="self.model.gpt_neox.layers.16.mlp.dense_h_to_4h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="10240, 1" offset="1419923154" size="20480" />
+		<layer id="1469" name="Constant_5820" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3951" name="self.model.gpt_neox.layers.16.mlp.dense_h_to_4h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
+		<layer id="1470" name="Constant_5822" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="1471" name="Constant_5824" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="2" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3952" name="self.model.gpt_neox.layers.16.mlp.dense_h_to_4h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="1472" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Slice_4" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="5" precision="FP32" names="1303">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3953" name="__module.model.gpt_neox.layers.16.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
+		<layer id="1473" name="__module.model.gpt_neox.layers.8.attention/aten::cat/Concat_1" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>10240</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="1341">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3954" name="Constant_299231_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 10240" offset="1419943634" size="20480" />
+		<layer id="1474" name="Constant_52743" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="FP32" />
+			</output>
+		</layer>
+		<layer id="1475" name="Constant_52736" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3955" name="Constant_299231" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+		<layer id="1476" name="Constant_52738" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>10240</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="1477" name="Constant_52740" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3956" name="__module.model.gpt_neox.layers.16.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1478" name="Concat_52741" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="FP32">
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
 					<dim>1</dim>
-					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2830">
-					<dim>-1</dim>
+				<port id="4" precision="I64">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1479" name="Broadcast_52744" type="Broadcast" version="opset3">
+			<data mode="numpy" />
+			<input>
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
+					<dim>4</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3957" name="__module.model.gpt_neox.layers.16.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
-			<data approximation_mode="ERF" />
+		<layer id="1480" name="ReadValue_50303" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.8.keypresent.8.key" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="2831">
+				<port id="1" precision="FP32" names="past_key_values.8.key">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3958" name="self.model.gpt_neox.layers.16.mlp.dense_4h_to_h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 10240" offset="1419964114" size="26214400" />
+		<layer id="1481" name="Constant_47826" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="3959" name="Convert_22094" type="Convert" version="opset1">
-			<data destination_type="f16" />
+		<layer id="1482" name="Gather_47827" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
+				<port id="1" precision="I32">
+					<dim>-1</dim>
+				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="3" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3960" name="self.model.gpt_neox.layers.16.mlp.dense_4h_to_h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="1446178514" size="2560" />
+		<layer id="1483" name="Constant_5774" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3961" name="Convert_22097" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
+		<layer id="1484" name="Constant_5776" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="1485" name="Constant_5778" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3962" name="self.model.gpt_neox.layers.16.mlp.dense_4h_to_h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1486" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Slice_5" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3963" name="self.model.gpt_neox.layers.16.mlp.dense_4h_to_h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="1446181074" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
+				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="3964" name="self.model.gpt_neox.layers.16.mlp.dense_4h_to_h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="3" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="5" precision="FP32" names="1296">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3965" name="self.model.gpt_neox.layers.16.mlp.dense_4h_to_h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
+		<layer id="1487" name="Constant_5787" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="I64" names="1297">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3966" name="__module.model.gpt_neox.layers.16.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
+		<layer id="1488" name="__module.model.gpt_neox.layers.8.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="1298,key.33">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3967" name="Constant_299232_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1446186194" size="5120" />
+		<layer id="1489" name="Constant_5830" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="FP16">
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1490" name="Constant_5832" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1491" name="Constant_5834" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3968" name="Constant_299232" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="1492" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Slice_6" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
+				<port id="3" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="5" precision="FP32" names="1304,k.17">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3969" name="__module.model.gpt_neox.layers.16.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1">
+		<layer id="1493" name="__module.model.gpt_neox.layers.8.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
+					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2834,input.69">
+				<port id="2" precision="FP32" names="1324">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3970" name="Constant_299233_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1446191314" size="5120" />
+		<layer id="1494" name="Constant_6111" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3971" name="Constant_299233" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+		<layer id="1495" name="Constant_6113" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="1496" name="Constant_6115" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3972" name="__module.model.gpt_neox.layers.16.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1497" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Slice_7" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="FP32">
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="5" precision="FP32" names="1334,x2.35">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3973" name="Constant_299234_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1446196434" size="5120" />
+		<layer id="1498" name="Constant_37358" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3974" name="Constant_299234" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3975" name="__module.model.gpt_neox.layers.16.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+		<layer id="1499" name="__module.model.gpt_neox.layers.8.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2700">
+				<port id="2" precision="FP32" names="1335">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3976" name="self.model.gpt_neox.layers.16.attention.query_key_value.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 2560" offset="1446201554" size="19660800" />
+		<layer id="1500" name="Constant_6008" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3977" name="Convert_22064" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="1501" name="Constant_6052" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3978" name="self.model.gpt_neox.layers.16.attention.query_key_value.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 1" offset="1465862354" size="7680" />
+		<layer id="1502" name="Constant_6054" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3979" name="Convert_22067" type="Convert" version="opset1">
-			<data destination_type="f16" />
+		<layer id="1503" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Slice_8" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
+				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="3980" name="self.model.gpt_neox.layers.16.attention.query_key_value.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="3" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3981" name="self.model.gpt_neox.layers.16.attention.query_key_value.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="7680, 1" offset="1465870034" size="15360" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>1</dim>
+				<port id="5" precision="FP32" names="1329">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3982" name="self.model.gpt_neox.layers.16.attention.query_key_value.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1504" name="__module.model.gpt_neox.layers.8.attention/aten::cat/Concat_2" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>1</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32" names="1337">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3983" name="self.model.gpt_neox.layers.16.attention.query_key_value.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="1505" name="__module.model.gpt_neox.layers.8.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32" names="1338">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3984" name="__module.model.gpt_neox.layers.16.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
+		<layer id="1506" name="__module.model.gpt_neox.layers.8.attention/aten::add/Add_1" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>7680</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="1339">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3985" name="Constant_299235_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 7680" offset="1465885394" size="15360" />
+		<layer id="1507" name="Constant_5840" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3986" name="Constant_299235" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+		<layer id="1508" name="Constant_5842" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>7680</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="1509" name="Constant_5844" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3987" name="__module.model.gpt_neox.layers.16.attention.query_key_value/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1510" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Slice_9" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>7680</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2707,qkv.65">
+				<port id="5" precision="FP32" names="1305">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>60</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3988" name="__module.model.gpt_neox.layers.16.attention/prim::ListConstruct/Concat" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080154" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1511" name="__module.model.gpt_neox.layers.8.attention/aten::cat/Concat_3" type="Concat" version="opset1">
+			<data axis="-1" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="1343">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3989" name="__module.model.gpt_neox.layers.16.attention/aten::view/Reshape" type="Reshape" version="opset1">
-			<data special_zero="true" />
+		<layer id="1512" name="__module.model.gpt_neox.layers.8.attention/aten::cat/Concat_4" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2711,qkv.67">
-					<dim>-1</dim>
+				<port id="2" precision="FP32" names="present.8.key">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>240</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3990" name="Constant_275029" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1513" name="Constant_52755" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
+			<output>
+				<port id="0" precision="FP32" />
+			</output>
+		</layer>
+		<layer id="1514" name="Constant_52748" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3991" name="Constant_275032" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1515" name="Constant_52750" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3992" name="Constant_275035" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1516" name="Constant_52752" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3993" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1517" name="Concat_52753" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2712">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
+				<port id="4" precision="I64">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="3994" name="Constant_9519" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="2713">
+		<layer id="1518" name="Broadcast_52756" type="Broadcast" version="opset3">
+			<data mode="numpy" />
+			<input>
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="3995" name="__module.model.gpt_neox.layers.16.attention/aten::permute/Transpose" type="Transpose" version="opset1">
+		<layer id="1519" name="ReadValue_50305" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.8.valuepresent.8.value" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
+					<dim>0</dim>
 					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2714,query.97">
+				<port id="1" precision="FP32" names="past_key_values.8.value">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -61064,41 +22956,14 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="3996" name="Constant_275041" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3997" name="Constant_275044" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="3998" name="Constant_275047" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1520" name="Constant_47829" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="3999" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1521" name="Gather_47830" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -61106,98 +22971,45 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="I32">
+					<dim>-1</dim>
 				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2721,q.33">
+				<port id="3" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4000" name="self.model.gpt_neox.layers.16.attention.rotary_emb.cos_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201080346" size="81920" />
-			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.16.attention.rotary_emb.cos_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4001" name="self.model.gpt_neox.layers.16.attention.rotary_emb.cos_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4002" name="__module.model.gpt_neox.layers.16.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1522" name="Constant_5792" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4003" name="Constant_275053" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4004" name="Constant_275056" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1523" name="Constant_5794" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4005" name="Constant_275059" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1524" name="Constant_5796" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4006" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1525" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Slice_10" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -61206,17 +23018,20 @@
 					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2715">
+				<port id="5" precision="FP32" names="1299">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -61224,15 +23039,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4007" name="Constant_9533" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="1526" name="Constant_5805" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64" names="2716">
+				<port id="0" precision="I64" names="1300">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4008" name="__module.model.gpt_neox.layers.16.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
+		<layer id="1527" name="__module.model.gpt_neox.layers.8.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -61245,7 +23060,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2717,key.97">
+				<port id="2" precision="FP32" names="1301">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -61253,8 +23068,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4009" name="__module.model.gpt_neox.layers.16.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="1528" name="__module.model.gpt_neox.layers.8.attention/aten::cat/Concat_5" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -61262,82 +23077,108 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="present.8.value">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4010" name="__module.model.gpt_neox.layers.16.attention/aten::size/Constant_2" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1529" name="Constant_6187" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4011" name="__module.model.gpt_neox.layers.16.attention/aten::size/Gather_2" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="1530" name="ShapeOf_39280" type="ShapeOf" version="opset3">
+			<data output_type="i64" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="2725" />
+				<port id="1" precision="I64">
+					<dim>4</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4012" name="Constant_337845_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
+		<layer id="1531" name="Constant_39281" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116168" size="8" />
 			<output>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4013" name="Constant_337845" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
+		<layer id="1532" name="Constant_39282" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="1" precision="FP32" />
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="4014" name="Constant_337838" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
-			<output>
+		<layer id="1533" name="Gather_39283" type="Gather" version="opset8">
+			<data batch_dims="0" />
+			<input>
 				<port id="0" precision="I64">
+					<dim>4</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64" />
+			</input>
+			<output>
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4015" name="Constant_337840" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<output>
+		<layer id="1534" name="Add_39286" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="I64" names="1348">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4016" name="Constant_337842" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
+		<layer id="1535" name="Constant_6190" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4017" name="Concat_337843" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="1536" name="__module.model.gpt_neox.layers.8.attention/aten::slice/Slice_14" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
 					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -61348,42 +23189,49 @@
 				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="5" precision="FP32" names="1352">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4018" name="Broadcast_337846" type="Broadcast" version="opset3">
-			<data mode="numpy" />
+		<layer id="1537" name="__module.model.gpt_neox.layers.8.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+			<data causal="false" />
 			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
+				<port id="1" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>0</dim>
+					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="4019" name="ReadValue_330457" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.16.keypresent.16.key" variable_type="f32" variable_shape="?,32,?,80" />
-			<input>
-				<port id="0" precision="FP32">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>0</dim>
+					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
+				<port id="3" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="past_key_values.16.key">
+				<port id="4" precision="FP32" names="1353,attn_output.49">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -61391,14 +23239,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4020" name="Constant_323481" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
+		<layer id="1538" name="__module.model.gpt_neox.layers.8.attention/aten::transpose/Constant" type="Const" version="opset1">
+			<data element_type="i32" shape="4" offset="201116200" size="16" />
 			<output>
-				<port id="0" precision="I64" />
+				<port id="0" precision="I32">
+					<dim>4</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4021" name="Gather_323482" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="1539" name="__module.model.gpt_neox.layers.8.attention/aten::transpose/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -61407,560 +23256,681 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I32">
-					<dim>-1</dim>
+					<dim>4</dim>
 				</port>
-				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="3" precision="FP32">
+				<port id="2" precision="FP32" names="1354">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4022" name="__module.model.gpt_neox.layers.16.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="1540" name="Constant_40992" type="Const" version="opset1">
+			<data element_type="i64" shape="3" offset="201116216" size="24" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>3</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1541" name="__module.model.gpt_neox.layers.8.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
+			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>80</dim>
 				</port>
+				<port id="1" precision="I64">
+					<dim>3</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="1357">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4023" name="__module.model.gpt_neox.layers.16.attention/aten::size/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1542" name="self.model.gpt_neox.layers.8.attention.dense.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 2560" offset="831674960" size="6553600" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4024" name="__module.model.gpt_neox.layers.16.attention/aten::size/Gather_3" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="1543" name="Convert_75264" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1544" name="self.model.gpt_neox.layers.8.attention.dense.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="838228560" size="2560" />
+			<output>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1545" name="Convert_75267" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="2727" />
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4025" name="__module.model.gpt_neox.layers.16.attention/aten::add_/Add" type="Add" version="opset1">
+		<layer id="1546" name="self.model.gpt_neox.layers.8.attention.dense.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64" names="2726,2730,2731,seq_len.65" />
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4026" name="__module.model.gpt_neox.layers.16.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1547" name="self.model.gpt_neox.layers.8.attention.dense.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="838231120" size="5120" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4027" name="__module.model.gpt_neox.layers.16.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1">
+		<layer id="1548" name="self.model.gpt_neox.layers.8.attention.dense.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4028" name="__module.model.gpt_neox.layers.16.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1549" name="self.model.gpt_neox.layers.8.attention.dense.weight/fq_weights_1/convert" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4029" name="__module.model.gpt_neox.layers.16.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1550" name="__module.model.gpt_neox.layers.8.attention.dense/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="2" precision="I64">
+			</input>
+			<output>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1551" name="Constant_37359" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="838236240" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
 					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
+			</output>
+		</layer>
+		<layer id="1552" name="__module.model.gpt_neox.layers.8.attention.dense/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
 					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2734,2737,cos.65">
+				<port id="2" precision="FP32" names="1360,1362,input.33">
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4030" name="Constant_38293" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
+		<layer id="1553" name="__module.model.gpt_neox.layers.8/aten::add/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I32" />
+				<port id="2" precision="FP32" names="1380">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4031" name="__module.model.gpt_neox.layers.16.attention/aten::index/Gather" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="1554" name="__module.model.gpt_neox.layers.8/aten::add/Add_1" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I32">
+				<port id="1" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="2740">
+				<port id="2" precision="FP32" names="1381,1383,hidden_states.39">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1555" name="__module.model.gpt_neox.layers.9.input_layernorm/aten::layer_norm/Multiply" type="Const" version="opset1">
+			<data element_type="i32" shape="1" offset="128867668" size="4" />
+			<output>
+				<port id="0" precision="I32">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4032" name="__module.model.gpt_neox.layers.16.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1">
+		<layer id="1556" name="__module.model.gpt_neox.layers.9.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
+			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="I32">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2741,cos.67">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>1</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1557" name="Constant_37362" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="838246480" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4033" name="__module.model.gpt_neox.layers.16.attention/aten::mul/Multiply" type="Multiply" version="opset1">
+		<layer id="1558" name="__module.model.gpt_neox.layers.9.input_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2745">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4034" name="Constant_275064" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1559" name="Constant_37363" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="838256720" size="10240" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1560" name="__module.model.gpt_neox.layers.9.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="1393,hidden_states.37">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4035" name="Constant_275063" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1561" name="self.model.gpt_neox.layers.9.attention.query_key_value.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 2560" offset="838266960" size="19660800" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4036" name="__module.model.gpt_neox.layers.16.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="1562" name="Convert_74923" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4037" name="__module.model.gpt_neox.layers.16.attention/aten::size/Constant_5" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1563" name="self.model.gpt_neox.layers.9.attention.query_key_value.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 1" offset="857927760" size="7680" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4038" name="__module.model.gpt_neox.layers.16.attention/aten::size/Gather_5" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="1564" name="Convert_74926" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="2746,2751" />
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4039" name="__module.model.gpt_neox.layers.16.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
+		<layer id="1565" name="self.model.gpt_neox.layers.9.attention.query_key_value.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64" />
+				<port id="2" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1566" name="self.model.gpt_neox.layers.9.attention.query_key_value.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="7680, 1" offset="857935440" size="15360" />
+			<output>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4040" name="__module.model.gpt_neox.layers.16.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1">
+		<layer id="1567" name="self.model.gpt_neox.layers.9.attention.query_key_value.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64" />
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64" names="2748,2749,2753,2754" />
+				<port id="2" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4041" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Constant_1" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1568" name="self.model.gpt_neox.layers.9.attention.query_key_value.weight/fq_weights_1/convert" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I32" />
+				<port id="1" precision="FP32">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4042" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1">
+		<layer id="1569" name="__module.model.gpt_neox.layers.9.attention.query_key_value/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4043" name="Constant_275062" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1570" name="Constant_37364" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 7680" offset="857950800" size="30720" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="FP32">
+					<dim>1</dim>
 					<dim>1</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4044" name="ScatterUpdate_275065" type="ScatterUpdate" version="opset3">
+		<layer id="1571" name="__module.model.gpt_neox.layers.9.attention.query_key_value/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>7680</dim>
 				</port>
-				<port id="2" precision="I64">
+				<port id="1" precision="FP32">
 					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
 					<dim>1</dim>
+					<dim>7680</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4045" name="Constant_275068" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="1400,qkv.37">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4046" name="Constant_275071" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1572" name="__module.model.gpt_neox.layers.9.attention/prim::ListConstruct" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201115992" size="32" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4047" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1573" name="__module.model.gpt_neox.layers.9.attention/aten::view/Reshape" type="Reshape" version="opset1">
+			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>7680</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2755,x2.65">
+				<port id="2" precision="FP32" names="1404,qkv.39">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4048" name="Constant_299236_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
+		<layer id="1574" name="Constant_6422" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4049" name="Constant_299236" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+		<layer id="1575" name="Constant_6424" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="1576" name="Constant_6426" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4050" name="__module.model.gpt_neox.layers.16.attention/aten::neg/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1577" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Slice_10" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
-				<port id="1" precision="FP32">
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2756">
+				<port id="5" precision="FP32" names="1411">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4051" name="Constant_275116" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4052" name="Constant_275121" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1578" name="Constant_6435" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="I64" names="1412">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4053" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1579" name="__module.model.gpt_neox.layers.9.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2750">
+				<port id="2" precision="FP32" names="1413">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4054" name="__module.model.gpt_neox.layers.16.attention/aten::cat/Concat" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="1580" name="__module.model.gpt_neox.layers.9.attention/aten::cat/Concat_5" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2758">
+				<port id="2" precision="FP32" names="present.9.value">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4055" name="self.model.gpt_neox.layers.16.attention.rotary_emb.sin_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201162356" size="81920" />
+		<layer id="1582" name="Constant_52767" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.16.attention.rotary_emb.sin_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
+				<port id="0" precision="FP32" />
 			</output>
 		</layer>
-		<layer id="4056" name="self.model.gpt_neox.layers.16.attention.rotary_emb.sin_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</input>
+		<layer id="1583" name="Constant_52760" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4057" name="__module.model.gpt_neox.layers.16.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1584" name="Constant_52762" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4058" name="__module.model.gpt_neox.layers.16.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1585" name="Constant_52764" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4059" name="__module.model.gpt_neox.layers.16.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1586" name="Concat_52765" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -61973,186 +23943,152 @@
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2735,2738,sin.65">
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="4" precision="I64">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4060" name="Constant_38296" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="4061" name="__module.model.gpt_neox.layers.16.attention/aten::index/Gather_1" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="1587" name="Broadcast_52768" type="Broadcast" version="opset3">
+			<data mode="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="2743">
-					<dim>-1</dim>
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4062" name="__module.model.gpt_neox.layers.16.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1">
+		<layer id="1588" name="ReadValue_50307" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.9.keypresent.9.key" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2744,sin.67">
+				<port id="1" precision="FP32" names="past_key_values.9.key">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4063" name="__module.model.gpt_neox.layers.16.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
+		<layer id="1589" name="Constant_47832" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="2" precision="FP32" names="2759">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="4064" name="__module.model.gpt_neox.layers.16.attention/aten::add/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1590" name="Gather_47833" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
+				<port id="1" precision="I32">
 					<dim>-1</dim>
-					<dim>20</dim>
 				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2760">
+				<port id="3" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4065" name="Constant_275167" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1591" name="Constant_6404" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4066" name="Constant_275170" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1592" name="Constant_6406" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4067" name="Constant_275173" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1593" name="Constant_6408" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4068" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1594" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Slice_5" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2722">
+				<port id="5" precision="FP32" names="1408">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>60</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4069" name="__module.model.gpt_neox.layers.16.attention/aten::cat/Concat_1" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="1595" name="Constant_6417" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
+			<output>
+				<port id="0" precision="I64" names="1409">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1596" name="__module.model.gpt_neox.layers.9.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2778,2788,query.101,query.99">
+				<port id="2" precision="FP32" names="1410,key.37">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -62160,41 +24096,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4070" name="Constant_275179" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1597" name="Constant_6460" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4071" name="Constant_275182" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1598" name="Constant_6462" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4072" name="Constant_275185" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1599" name="Constant_6464" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4073" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1600" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Slice_6" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -62203,42 +24129,20 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="2723,k.33">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4074" name="__module.model.gpt_neox.layers.16.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2761">
+				<port id="5" precision="FP32" names="1416,k.19">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -62246,30 +24150,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4075" name="Constant_275189" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4076" name="Constant_275188" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4077" name="__module.model.gpt_neox.layers.16.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="1601" name="__module.model.gpt_neox.layers.9.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -62277,129 +24159,47 @@
 					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4078" name="__module.model.gpt_neox.layers.16.attention/aten::size/Constant_7" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="4079" name="__module.model.gpt_neox.layers.16.attention/aten::size/Gather_7" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="2762,2767" />
-			</output>
-		</layer>
-		<layer id="4080" name="__module.model.gpt_neox.layers.16.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" />
-			</output>
-		</layer>
-		<layer id="4081" name="__module.model.gpt_neox.layers.16.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-			</input>
-			<output>
-				<port id="1" precision="I64" names="2764,2765,2769,2770" />
-			</output>
-		</layer>
-		<layer id="4082" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="4083" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4084" name="Constant_275187" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4085" name="ScatterUpdate_275190" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
 					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="1436">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4086" name="Constant_275193" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1602" name="Constant_6741" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4087" name="Constant_275196" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1603" name="Constant_6743" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1604" name="Constant_6745" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4088" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1605" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Slice_7" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -62408,17 +24208,20 @@
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2771,x2.67">
+				<port id="5" precision="FP32" names="1446,x2.39">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -62426,32 +24229,10 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4089" name="Constant_299237_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4090" name="Constant_299237" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
+		<layer id="1606" name="Constant_37366" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>1</dim>
@@ -62459,7 +24240,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4091" name="__module.model.gpt_neox.layers.16.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
+		<layer id="1607" name="__module.model.gpt_neox.layers.9.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -62476,7 +24257,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2772">
+				<port id="2" precision="FP32" names="1447">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -62484,30 +24265,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4092" name="Constant_275241" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1608" name="Constant_6638" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4093" name="Constant_275246" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1609" name="Constant_6682" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1610" name="Constant_6684" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4094" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1611" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Slice_8" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -62516,17 +24298,20 @@
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2766">
+				<port id="5" precision="FP32" names="1441">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -62534,7 +24319,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4095" name="__module.model.gpt_neox.layers.16.attention/aten::cat/Concat_2" type="Concat" version="opset1">
+		<layer id="1612" name="__module.model.gpt_neox.layers.9.attention/aten::cat/Concat_2" type="Concat" version="opset1">
 			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
@@ -62551,7 +24336,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2774">
+				<port id="2" precision="FP32" names="1449">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -62559,7 +24344,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4096" name="__module.model.gpt_neox.layers.16.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
+		<layer id="1613" name="__module.model.gpt_neox.layers.9.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -62576,7 +24361,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2775">
+				<port id="2" precision="FP32" names="1450">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -62584,7 +24369,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4097" name="__module.model.gpt_neox.layers.16.attention/aten::add/Add_1" type="Add" version="opset1">
+		<layer id="1614" name="__module.model.gpt_neox.layers.9.attention/aten::add/Add_1" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -62601,7 +24386,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2776">
+				<port id="2" precision="FP32" names="1451">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -62609,41 +24394,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4098" name="Constant_275291" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1615" name="Constant_6470" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4099" name="Constant_275294" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1616" name="Constant_6472" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4100" name="Constant_275297" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1617" name="Constant_6474" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4101" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1618" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Slice_9" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -62652,17 +24427,20 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2724">
+				<port id="5" precision="FP32" names="1417">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -62670,7 +24448,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4102" name="__module.model.gpt_neox.layers.16.attention/aten::cat/Concat_3" type="Concat" version="opset1">
+		<layer id="1619" name="__module.model.gpt_neox.layers.9.attention/aten::cat/Concat_3" type="Concat" version="opset1">
 			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
@@ -62687,7 +24465,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2780">
+				<port id="2" precision="FP32" names="1455">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -62695,7 +24473,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4103" name="__module.model.gpt_neox.layers.16.attention/aten::to/Convert_1" type="Concat" version="opset1">
+		<layer id="1620" name="__module.model.gpt_neox.layers.9.attention/aten::cat/Concat_4" type="Concat" version="opset1">
 			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
@@ -62712,7 +24490,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="present.16.key">
+				<port id="2" precision="FP32" names="present.9.key">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -62720,49 +24498,37 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4104" name="Constant_337858_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="4105" name="Constant_337858" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
+		<layer id="1632" name="Constant_52779" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="1" precision="FP32" />
+				<port id="0" precision="FP32" />
 			</output>
 		</layer>
-		<layer id="4106" name="Constant_337851" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
+		<layer id="1633" name="Constant_52772" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4107" name="Constant_337853" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
+		<layer id="1634" name="Constant_52774" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4108" name="Constant_337855" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
+		<layer id="1635" name="Constant_52776" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4109" name="Concat_337856" type="Concat" version="opset1">
+		<layer id="1636" name="Concat_52777" type="Concat" version="opset1">
 			<data axis="0" />
 			<input>
 				<port id="0" precision="I64">
@@ -62784,7 +24550,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4110" name="Broadcast_337859" type="Broadcast" version="opset3">
+		<layer id="1637" name="Broadcast_52780" type="Broadcast" version="opset3">
 			<data mode="numpy" />
 			<input>
 				<port id="0" precision="FP32" />
@@ -62801,8 +24567,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4111" name="ReadValue_330459" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.16.valuepresent.16.value" variable_type="f32" variable_shape="?,32,?,80" />
+		<layer id="1638" name="ReadValue_50285" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.31.valuepresent.31.value" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -62812,7 +24578,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="past_key_values.16.value">
+				<port id="1" precision="FP32" names="past_key_values.31.value">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -62820,13 +24586,13 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4112" name="Constant_323484" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
+		<layer id="1639" name="Constant_47967" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="4113" name="Gather_323485" type="Gather" version="opset8">
+		<layer id="1640" name="Gather_47968" type="Gather" version="opset8">
 			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="FP32">
@@ -62849,41 +24615,468 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4114" name="Constant_275303" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
+		<layer id="1641" name="Constant_37360" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="857981520" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1642" name="__module.model.gpt_neox.layers.9.post_attention_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1643" name="Constant_37361" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="857991760" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1644" name="__module.model.gpt_neox.layers.9.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="1481">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1645" name="self.model.gpt_neox.layers.9.mlp.dense_h_to_4h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 2560" offset="858002000" size="26214400" />
+			<output>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1646" name="Convert_74364" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1647" name="self.model.gpt_neox.layers.9.mlp.dense_h_to_4h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 1" offset="884216400" size="10240" />
+			<output>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1648" name="Convert_74367" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1649" name="self.model.gpt_neox.layers.9.mlp.dense_h_to_4h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1650" name="self.model.gpt_neox.layers.9.mlp.dense_h_to_4h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="10240, 1" offset="884226640" size="20480" />
+			<output>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1651" name="self.model.gpt_neox.layers.9.mlp.dense_h_to_4h.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1652" name="__module.model.gpt_neox.layers.9.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<rt_info>
-				<attribute name="precise" version="0" />
+				<attribute name="decompression" version="0" />
 			</rt_info>
+			<input>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1653" name="__module.model.gpt_neox.layers.9.mlp.dense_h_to_4h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1654" name="self.model.gpt_neox.layers.9.mlp.dense_h_to_4h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="10240" offset="884247120" size="20480" />
+			<output>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.9.mlp.dense_h_to_4h.bias">
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1655" name="__module.model.gpt_neox.layers.9.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<rt_info>
+				<attribute name="decompression" version="0" />
+			</rt_info>
+			<input>
+				<port id="0" precision="BF16">
+					<dim>10240</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1656" name="__module.model.gpt_neox.layers.9.mlp.dense_h_to_4h/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="1486">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1657" name="__module.model.gpt_neox.layers.9.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
+			<data approximation_mode="ERF" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP32" names="1487">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1658" name="self.model.gpt_neox.layers.9.mlp.dense_4h_to_h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 10240" offset="884267600" size="26214400" />
+			<output>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1659" name="Convert_74374" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1660" name="self.model.gpt_neox.layers.9.mlp.dense_4h_to_h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="910482000" size="2560" />
+			<output>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1661" name="Convert_74377" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1662" name="self.model.gpt_neox.layers.9.mlp.dense_4h_to_h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1663" name="self.model.gpt_neox.layers.9.mlp.dense_4h_to_h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="910484560" size="5120" />
+			<output>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1664" name="self.model.gpt_neox.layers.9.mlp.dense_4h_to_h.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1665" name="__module.model.gpt_neox.layers.9.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<rt_info>
+				<attribute name="decompression" version="0" />
+			</rt_info>
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1666" name="__module.model.gpt_neox.layers.9.mlp.dense_4h_to_h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1667" name="self.model.gpt_neox.layers.9.mlp.dense_4h_to_h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="2560" offset="910489680" size="5120" />
+			<output>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.9.mlp.dense_4h_to_h.bias">
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4115" name="Constant_275306" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
+		<layer id="1668" name="__module.model.gpt_neox.layers.9.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<rt_info>
-				<attribute name="precise" version="0" />
+				<attribute name="decompression" version="0" />
 			</rt_info>
+			<input>
+				<port id="0" precision="BF16">
+					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1669" name="__module.model.gpt_neox.layers.9.mlp.dense_4h_to_h/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="1490,input.39">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1670" name="Constant_6386" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1671" name="Constant_6388" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4116" name="Constant_275309" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1672" name="Constant_6390" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4117" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1673" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Slice" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -62892,17 +25085,20 @@
 					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2718">
+				<port id="5" precision="FP32" names="1405">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -62910,15 +25106,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4118" name="Constant_9547" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="1674" name="Constant_6399" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64" names="2719">
+				<port id="0" precision="I64" names="1406">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4119" name="__module.model.gpt_neox.layers.16.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
+		<layer id="1675" name="__module.model.gpt_neox.layers.9.attention/aten::permute/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -62931,32 +25127,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2720">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4120" name="__module.model.gpt_neox.layers.16.attention/aten::cat/Concat_5" type="Concat" version="opset1">
-			<data axis="-2" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="present.16.value">
+				<port id="2" precision="FP32" names="1407,query.19">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -62964,41 +25135,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4121" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Slice_11" type="Const" version="opset1">
-			<data element_type="boolean" shape="1, 1, 2048, 2048" offset="201244276" size="4194304" />
+		<layer id="1676" name="Constant_6440" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="BOOL" names="2797">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4122" name="Constant_275313" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1677" name="Constant_6442" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>3</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4123" name="Constant_275312" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162298" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1678" name="Constant_6444" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4124" name="__module.model.gpt_neox.layers.16.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="1679" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Slice_1" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -63006,118 +25167,84 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-			</input>
-			<output>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="4125" name="__module.model.gpt_neox.layers.16.attention/aten::size/Constant_8" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="4126" name="__module.model.gpt_neox.layers.16.attention/aten::size/Gather_8" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="2792" />
+				<port id="5" precision="FP32" names="1414,q.19">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4127" name="__module.model.gpt_neox.layers.16.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="1680" name="__module.model.gpt_neox.layers.9.attention/aten::mul/Multiply" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="1420">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4128" name="__module.model.gpt_neox.layers.16.attention/aten::size/Constant_9" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1681" name="Constant_6598" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="4129" name="__module.model.gpt_neox.layers.16.attention/aten::size/Gather_9" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="2790" />
-			</output>
-		</layer>
-		<layer id="4130" name="__module.model.gpt_neox.layers.16.attention/aten::sub/Subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" names="2794,2795" />
-			</output>
-		</layer>
-		<layer id="4131" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Constant_4" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
 			</output>
 		</layer>
-		<layer id="4132" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="1682" name="Constant_6600" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4133" name="Constant_275311" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1683" name="Constant_6602" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4134" name="ScatterUpdate_275314" type="ScatterUpdate" version="opset3">
+		<layer id="1684" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Slice_2" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -63125,146 +25252,89 @@
 				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
 				<port id="4" precision="I64">
-					<dim>3</dim>
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="4135" name="Constant_275315" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="5" precision="FP32" names="1430,x2.37">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4136" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="1685" name="Constant_37365" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4137" name="ScatterUpdate_275316" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
 					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
 					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4138" name="Constant_275319" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438612" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
 			</output>
 		</layer>
-		<layer id="4139" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1686" name="__module.model.gpt_neox.layers.9.attention/aten::neg/Multiply" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="BOOL">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+				<port id="1" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>3</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="BOOL" names="2798">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="4140" name="Constant_275392" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="1431">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4141" name="Constant_275393" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1687" name="Constant_6495" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4142" name="Constant_275389" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1688" name="Constant_6539" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4143" name="Constant_275388" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1689" name="Constant_6541" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4144" name="ScatterUpdate_275394" type="ScatterUpdate" version="opset3">
+		<layer id="1690" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Slice_3" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -63272,133 +25342,192 @@
 				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="5" precision="FP32" names="1425">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4145" name="Constant_275397" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1691" name="__module.model.gpt_neox.layers.9.attention/aten::cat/Concat" type="Concat" version="opset1">
+			<data axis="-1" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="1433">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4146" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1692" name="__module.model.gpt_neox.layers.9.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
 					<dim>1</dim>
 					<dim>-1</dim>
-					<dim>2048</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="1434">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+			</output>
+		</layer>
+		<layer id="1693" name="__module.model.gpt_neox.layers.9.attention/aten::add/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="BOOL" names="2799,2800">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="1435">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4147" name="__module.model.gpt_neox.layers.16.attention/aten::where/Convert_1_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
+		<layer id="1694" name="Constant_6450" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4148" name="__module.model.gpt_neox.layers.16.attention/aten::where/Convert_1" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
+		<layer id="1695" name="Constant_6452" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
-				<port id="1" precision="FP32" />
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4149" name="__module.model.gpt_neox.layers.16.attention/aten::full/Convert" type="Const" version="opset1">
-			<data element_type="f32" shape="" offset="205438636" size="4" />
+		<layer id="1696" name="Constant_6454" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="FP32" names="2787" />
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4150" name="__module.model.gpt_neox.layers.16.attention/aten::where/Select" type="Select" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1697" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Slice_4" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32" />
-				<port id="2" precision="FP32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="2801,causal_mask.65">
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="5" precision="FP32" names="1415">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4151" name="Add_290434" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1698" name="__module.model.gpt_neox.layers.9.attention/aten::cat/Concat_1" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="1453">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1699" name="Constant_6817" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4152" name="ShapeOf_290435" type="ShapeOf" version="opset3">
+		<layer id="1700" name="ShapeOf_39351" type="ShapeOf" version="opset3">
 			<data output_type="i64" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
@@ -63407,27 +25536,21 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4153" name="Constant_254569" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1701" name="Constant_39352" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116168" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4154" name="__module.model.gpt_neox.layers.16.attention/aten::size/Constant_10" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1702" name="Constant_39353" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="4155" name="__module.model.gpt_neox.layers.16.attention/aten::size/Gather_10" type="Gather" version="opset8">
+		<layer id="1703" name="Gather_39354" type="Gather" version="opset8">
 			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="I64">
@@ -63436,16 +25559,16 @@
 				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I32" />
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="2785">
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4156" name="__module.model.gpt_neox.layers.16.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="1704" name="Add_39357" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="I64">
 					<dim>1</dim>
@@ -63453,49 +25576,22 @@
 				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64" names="2802">
-					<dim>4</dim>
+				<port id="2" precision="I64" names="1460">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4157" name="__module.model.gpt_neox.layers.16.attention/aten::expand/Abs" type="Abs" version="opset1">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
+		<layer id="1705" name="Constant_6820" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4158" name="Maximum_290436" type="Maximum" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4159" name="__module.model.gpt_neox.layers.16.attention/aten::add/Add_2" type="Broadcast" version="opset3">
-			<data mode="numpy" />
+		<layer id="1706" name="__module.model.gpt_neox.layers.9.attention/aten::slice/Slice_14" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -63504,19 +25600,28 @@
 					<dim>-1</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2804">
-					<dim>-1</dim>
+				<port id="5" precision="FP32" names="1464">
 					<dim>-1</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
 					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4160" name="__module.model.gpt_neox.layers.16.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+		<layer id="1707" name="__module.model.gpt_neox.layers.9.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
 			<data causal="false" />
 			<input>
 				<port id="0" precision="FP32">
@@ -63539,13 +25644,13 @@
 				</port>
 				<port id="3" precision="FP32">
 					<dim>-1</dim>
-					<dim>-1</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
 					<dim>-1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2805,2806,sdpa_result.33,tensor.65">
+				<port id="4" precision="FP32" names="1465,attn_output.55">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -63553,15 +25658,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4161" name="Constant_9988" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="1708" name="__module.model.gpt_neox.layers.9.attention/aten::transpose/Constant" type="Const" version="opset1">
+			<data element_type="i32" shape="4" offset="201116200" size="16" />
 			<output>
-				<port id="0" precision="I64" names="2807">
+				<port id="0" precision="I32">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4162" name="__module.model.gpt_neox.layers.16.attention/aten::permute/Transpose_3" type="Transpose" version="opset1">
+		<layer id="1709" name="__module.model.gpt_neox.layers.9.attention/aten::transpose/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -63569,12 +25674,12 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="I32">
 					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2808">
+				<port id="2" precision="FP32" names="1466">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -63582,18 +25687,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4163" name="__module.model.gpt_neox.layers.16.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438650" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1710" name="Constant_40993" type="Const" version="opset1">
+			<data element_type="i64" shape="3" offset="201116216" size="24" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>3</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4164" name="__module.model.gpt_neox.layers.16.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
+		<layer id="1711" name="__module.model.gpt_neox.layers.9.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
 			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -63607,15 +25709,15 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2813">
+				<port id="2" precision="FP32" names="1469">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4165" name="self.model.gpt_neox.layers.16.attention.dense.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 2560" offset="1465900754" size="6553600" />
+		<layer id="1712" name="self.model.gpt_neox.layers.9.attention.dense.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 2560" offset="910494800" size="6553600" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>2560</dim>
@@ -63623,7 +25725,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4166" name="Convert_22074" type="Convert" version="opset1">
+		<layer id="1713" name="Convert_75275" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -63638,8 +25740,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4167" name="self.model.gpt_neox.layers.16.attention.dense.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="1472454354" size="2560" />
+		<layer id="1714" name="self.model.gpt_neox.layers.9.attention.dense.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="917048400" size="2560" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>2560</dim>
@@ -63647,7 +25749,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4168" name="Convert_22077" type="Convert" version="opset1">
+		<layer id="1715" name="Convert_75278" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -63662,7 +25764,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4169" name="self.model.gpt_neox.layers.16.attention.dense.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="1716" name="self.model.gpt_neox.layers.9.attention.dense.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -63681,8 +25783,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4170" name="self.model.gpt_neox.layers.16.attention.dense.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="1472456914" size="5120" />
+		<layer id="1717" name="self.model.gpt_neox.layers.9.attention.dense.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="917050960" size="5120" />
 			<output>
 				<port id="0" precision="FP16">
 					<dim>2560</dim>
@@ -63690,7 +25792,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4171" name="self.model.gpt_neox.layers.16.attention.dense.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="1718" name="self.model.gpt_neox.layers.9.attention.dense.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -63709,11 +25811,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4172" name="self.model.gpt_neox.layers.16.attention.dense.weight" type="Convert" version="opset1">
+		<layer id="1719" name="self.model.gpt_neox.layers.9.attention.dense.weight/fq_weights_1/convert" type="Convert" version="opset1">
 			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
 			<input>
 				<port id="0" precision="FP16">
 					<dim>2560</dim>
@@ -63727,7 +25826,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4173" name="__module.model.gpt_neox.layers.16.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="1720" name="__module.model.gpt_neox.layers.9.attention.dense/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -63748,37 +25847,17 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4174" name="Constant_299238_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1472462034" size="5120" />
+		<layer id="1721" name="Constant_37367" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="917056080" size="10240" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4175" name="Constant_299238" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4176" name="__module.model.gpt_neox.layers.16.attention.dense/aten::linear/Add" type="Add" version="opset1">
+		<layer id="1722" name="__module.model.gpt_neox.layers.9.attention.dense/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -63793,14 +25872,14 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2816,2818,input.67">
+				<port id="2" precision="FP32" names="1472,1474,input.37">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4177" name="__module.model.gpt_neox.layers.16/aten::add/Add" type="Add" version="opset1">
+		<layer id="1723" name="__module.model.gpt_neox.layers.9/aten::add/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -63815,14 +25894,14 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2836">
+				<port id="2" precision="FP32" names="1492">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4178" name="__module.model.gpt_neox.layers.16/aten::add/Add_1" type="Add" version="opset1">
+		<layer id="1724" name="__module.model.gpt_neox.layers.9/aten::add/Add_1" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -63837,22 +25916,22 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2837,2839,hidden_states.35">
+				<port id="2" precision="FP32" names="1493,1495,hidden_states.43">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4179" name="__module.model.gpt_neox.layers.17.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867670" size="4" />
+		<layer id="1725" name="__module.model.gpt_neox.layers.10.input_layernorm/aten::layer_norm/Multiply" type="Const" version="opset1">
+			<data element_type="i32" shape="1" offset="128867668" size="4" />
 			<output>
 				<port id="0" precision="I32">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4180" name="__module.model.gpt_neox.layers.17.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
+		<layer id="1726" name="__module.model.gpt_neox.layers.10.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
 			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
 			<input>
 				<port id="0" precision="FP32">
@@ -63872,37 +25951,17 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4181" name="Constant_299239_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1472467154" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4182" name="Constant_299239" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="1727" name="Constant_37368" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="917066320" size="10240" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4183" name="__module.model.gpt_neox.layers.17.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
+		<layer id="1728" name="__module.model.gpt_neox.layers.10.post_attention_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -63924,37 +25983,17 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4184" name="Constant_299240_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1472472274" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4185" name="Constant_299240" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="1729" name="Constant_37369" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="917076560" size="10240" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4186" name="__module.model.gpt_neox.layers.17.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+		<layer id="1730" name="__module.model.gpt_neox.layers.10.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -63969,15 +26008,15 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2974">
+				<port id="2" precision="FP32" names="1593">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4187" name="self.model.gpt_neox.layers.17.mlp.dense_h_to_4h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 2560" offset="1472477394" size="26214400" />
+		<layer id="1731" name="self.model.gpt_neox.layers.10.mlp.dense_h_to_4h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 2560" offset="917086800" size="26214400" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>10240</dim>
@@ -63985,7 +26024,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4188" name="Convert_22124" type="Convert" version="opset1">
+		<layer id="1732" name="Convert_74384" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -64000,8 +26039,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4189" name="self.model.gpt_neox.layers.17.mlp.dense_h_to_4h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 1" offset="1498691794" size="10240" />
+		<layer id="1733" name="self.model.gpt_neox.layers.10.mlp.dense_h_to_4h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 1" offset="943301200" size="10240" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>10240</dim>
@@ -64009,7 +26048,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4190" name="Convert_22127" type="Convert" version="opset1">
+		<layer id="1734" name="Convert_74387" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -64024,7 +26063,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4191" name="self.model.gpt_neox.layers.17.mlp.dense_h_to_4h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="1735" name="self.model.gpt_neox.layers.10.mlp.dense_h_to_4h.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -64043,8 +26082,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4192" name="self.model.gpt_neox.layers.17.mlp.dense_h_to_4h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="10240, 1" offset="1498702034" size="20480" />
+		<layer id="1736" name="self.model.gpt_neox.layers.10.mlp.dense_h_to_4h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="10240, 1" offset="943311440" size="20480" />
 			<output>
 				<port id="0" precision="FP16">
 					<dim>10240</dim>
@@ -64052,7 +26091,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4193" name="self.model.gpt_neox.layers.17.mlp.dense_h_to_4h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="1737" name="self.model.gpt_neox.layers.10.mlp.dense_h_to_4h.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -64071,7 +26110,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4194" name="self.model.gpt_neox.layers.17.mlp.dense_h_to_4h.weight" type="Convert" version="opset1">
+		<layer id="1738" name="__module.model.gpt_neox.layers.10.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
@@ -64089,7 +26128,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4195" name="__module.model.gpt_neox.layers.17.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="1739" name="__module.model.gpt_neox.layers.10.mlp.dense_h_to_4h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -64110,37 +26149,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4196" name="Constant_299241_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 10240" offset="1498722514" size="20480" />
+		<layer id="1740" name="self.model.gpt_neox.layers.10.mlp.dense_h_to_4h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="10240" offset="943331920" size="20480" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.10.mlp.dense_h_to_4h.bias">
 					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4197" name="Constant_299241" type="Convert" version="opset1">
+		<layer id="1741" name="__module.model.gpt_neox.layers.10.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
 			</rt_info>
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16">
 					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4198" name="__module.model.gpt_neox.layers.17.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1">
+		<layer id="1742" name="__module.model.gpt_neox.layers.10.mlp.dense_h_to_4h/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -64149,20 +26182,18 @@
 					<dim>10240</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2979">
+				<port id="2" precision="FP32" names="1598">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4199" name="__module.model.gpt_neox.layers.17.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
+		<layer id="1743" name="__module.model.gpt_neox.layers.10.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
 			<data approximation_mode="ERF" />
 			<input>
 				<port id="0" precision="FP32">
@@ -64172,15 +26203,15 @@
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="2980">
+				<port id="1" precision="FP32" names="1599">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4200" name="self.model.gpt_neox.layers.17.mlp.dense_4h_to_h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 10240" offset="1498742994" size="26214400" />
+		<layer id="1744" name="self.model.gpt_neox.layers.10.mlp.dense_4h_to_h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 10240" offset="943352400" size="26214400" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>2560</dim>
@@ -64188,7 +26219,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4201" name="Convert_22134" type="Convert" version="opset1">
+		<layer id="1745" name="Convert_74394" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -64203,8 +26234,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4202" name="self.model.gpt_neox.layers.17.mlp.dense_4h_to_h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="1524957394" size="2560" />
+		<layer id="1746" name="self.model.gpt_neox.layers.10.mlp.dense_4h_to_h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="969566800" size="2560" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>2560</dim>
@@ -64212,7 +26243,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4203" name="Convert_22137" type="Convert" version="opset1">
+		<layer id="1747" name="Convert_74397" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -64227,7 +26258,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4204" name="self.model.gpt_neox.layers.17.mlp.dense_4h_to_h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="1748" name="self.model.gpt_neox.layers.10.mlp.dense_4h_to_h.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -64246,8 +26277,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4205" name="self.model.gpt_neox.layers.17.mlp.dense_4h_to_h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="1524959954" size="5120" />
+		<layer id="1749" name="self.model.gpt_neox.layers.10.mlp.dense_4h_to_h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="969569360" size="5120" />
 			<output>
 				<port id="0" precision="FP16">
 					<dim>2560</dim>
@@ -64255,7 +26286,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4206" name="self.model.gpt_neox.layers.17.mlp.dense_4h_to_h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="1750" name="self.model.gpt_neox.layers.10.mlp.dense_4h_to_h.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -64274,7 +26305,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4207" name="self.model.gpt_neox.layers.17.mlp.dense_4h_to_h.weight" type="Convert" version="opset1">
+		<layer id="1751" name="__module.model.gpt_neox.layers.10.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
@@ -64292,7 +26323,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4208" name="__module.model.gpt_neox.layers.17.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="1752" name="__module.model.gpt_neox.layers.10.mlp.dense_4h_to_h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -64313,37 +26344,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4209" name="Constant_299242_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1524965074" size="5120" />
+		<layer id="1753" name="self.model.gpt_neox.layers.10.mlp.dense_4h_to_h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="2560" offset="969574480" size="5120" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.10.mlp.dense_4h_to_h.bias">
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4210" name="Constant_299242" type="Convert" version="opset1">
+		<layer id="1754" name="__module.model.gpt_neox.layers.10.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
 			</rt_info>
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16">
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4211" name="__module.model.gpt_neox.layers.17.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1">
+		<layer id="1755" name="__module.model.gpt_neox.layers.10.mlp.dense_4h_to_h/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -64352,50 +26377,28 @@
 					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2983,input.73">
+				<port id="2" precision="FP32" names="1602,input.43">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4212" name="Constant_299243_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1524970194" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4213" name="Constant_299243" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="1756" name="Constant_37370" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="969579600" size="10240" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4214" name="__module.model.gpt_neox.layers.17.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
+		<layer id="1757" name="__module.model.gpt_neox.layers.10.input_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -64417,37 +26420,17 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4215" name="Constant_299244_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1524975314" size="5120" />
+		<layer id="1758" name="Constant_37371" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="969589840" size="10240" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4216" name="Constant_299244" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4217" name="__module.model.gpt_neox.layers.17.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+		<layer id="1759" name="__module.model.gpt_neox.layers.10.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -64462,15 +26445,15 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2849">
+				<port id="2" precision="FP32" names="1505,hidden_states.41">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4218" name="self.model.gpt_neox.layers.17.attention.query_key_value.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 2560" offset="1524980434" size="19660800" />
+		<layer id="1760" name="self.model.gpt_neox.layers.10.attention.query_key_value.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 2560" offset="969600080" size="19660800" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>7680</dim>
@@ -64478,7 +26461,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4219" name="Convert_22104" type="Convert" version="opset1">
+		<layer id="1761" name="Convert_74934" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -64493,8 +26476,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4220" name="self.model.gpt_neox.layers.17.attention.query_key_value.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 1" offset="1544641234" size="7680" />
+		<layer id="1762" name="self.model.gpt_neox.layers.10.attention.query_key_value.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 1" offset="989260880" size="7680" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>7680</dim>
@@ -64502,7 +26485,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4221" name="Convert_22107" type="Convert" version="opset1">
+		<layer id="1763" name="Convert_74937" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -64517,7 +26500,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4222" name="self.model.gpt_neox.layers.17.attention.query_key_value.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="1764" name="self.model.gpt_neox.layers.10.attention.query_key_value.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -64536,8 +26519,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4223" name="self.model.gpt_neox.layers.17.attention.query_key_value.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="7680, 1" offset="1544648914" size="15360" />
+		<layer id="1765" name="self.model.gpt_neox.layers.10.attention.query_key_value.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="7680, 1" offset="989268560" size="15360" />
 			<output>
 				<port id="0" precision="FP16">
 					<dim>7680</dim>
@@ -64545,7 +26528,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4224" name="self.model.gpt_neox.layers.17.attention.query_key_value.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="1766" name="self.model.gpt_neox.layers.10.attention.query_key_value.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -64564,11 +26547,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4225" name="self.model.gpt_neox.layers.17.attention.query_key_value.weight" type="Convert" version="opset1">
+		<layer id="1767" name="self.model.gpt_neox.layers.10.attention.query_key_value.weight/fq_weights_1/convert" type="Convert" version="opset1">
 			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
 			<input>
 				<port id="0" precision="FP16">
 					<dim>7680</dim>
@@ -64582,7 +26562,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4226" name="__module.model.gpt_neox.layers.17.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="1768" name="__module.model.gpt_neox.layers.10.attention.query_key_value/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -64603,37 +26583,17 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4227" name="Constant_299245_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 7680" offset="1544664274" size="15360" />
+		<layer id="1769" name="Constant_37372" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 7680" offset="989283920" size="30720" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4228" name="Constant_299245" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4229" name="__module.model.gpt_neox.layers.17.attention.query_key_value/aten::linear/Add" type="Add" version="opset1">
+		<layer id="1770" name="__module.model.gpt_neox.layers.10.attention.query_key_value/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -64648,25 +26608,22 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2856,qkv.69">
+				<port id="2" precision="FP32" names="1512,qkv.41">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4230" name="__module.model.gpt_neox.layers.17.attention/prim::ListConstruct/Concat" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080154" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1771" name="__module.model.gpt_neox.layers.10.attention/prim::ListConstruct" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201115992" size="32" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4231" name="__module.model.gpt_neox.layers.17.attention/aten::view/Reshape" type="Reshape" version="opset1">
+		<layer id="1772" name="__module.model.gpt_neox.layers.10.attention/aten::view/Reshape" type="Reshape" version="opset1">
 			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -64679,7 +26636,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2860,qkv.71">
+				<port id="2" precision="FP32" names="1516,qkv.43">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -64687,41 +26644,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4232" name="Constant_275404" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1773" name="Constant_7016" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4233" name="Constant_275407" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1774" name="Constant_7018" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4234" name="Constant_275410" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1775" name="Constant_7020" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4235" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1776" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Slice" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -64730,17 +26677,20 @@
 					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2861">
+				<port id="5" precision="FP32" names="1517">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -64748,15 +26698,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4236" name="Constant_10100" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="1777" name="Constant_7029" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64" names="2862">
+				<port id="0" precision="I64" names="1518">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4237" name="__module.model.gpt_neox.layers.17.attention/aten::permute/Transpose" type="Transpose" version="opset1">
+		<layer id="1778" name="__module.model.gpt_neox.layers.10.attention/aten::permute/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -64769,7 +26719,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2863,query.103">
+				<port id="2" precision="FP32" names="1519,query.21">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -64777,41 +26727,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4238" name="Constant_275416" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1779" name="Constant_7070" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4239" name="Constant_275419" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1780" name="Constant_7072" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4240" name="Constant_275422" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1781" name="Constant_7074" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4241" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1782" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Slice_1" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -64820,17 +26760,20 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2870,q.35">
+				<port id="5" precision="FP32" names="1526,q.21">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -64838,219 +26781,152 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4242" name="self.model.gpt_neox.layers.17.attention.rotary_emb.cos_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201080346" size="81920" />
-			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.17.attention.rotary_emb.cos_cached">
-					<dim>2048</dim>
+		<layer id="1783" name="__module.model.gpt_neox.layers.10.attention/aten::mul/Multiply" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="4243" name="self.model.gpt_neox.layers.17.attention.rotary_emb.cos_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
+				<port id="2" precision="FP32" names="1532">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4244" name="__module.model.gpt_neox.layers.17.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1784" name="Constant_7228" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4245" name="Constant_275428" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4246" name="Constant_275431" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1785" name="Constant_7230" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4247" name="Constant_275434" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1786" name="Constant_7232" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4248" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1787" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Slice_2" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>240</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2864">
-					<dim>-1</dim>
+				<port id="5" precision="FP32" names="1542,x2.41">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4249" name="Constant_10114" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="1788" name="Constant_37373" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="0" precision="I64" names="2865">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4250" name="__module.model.gpt_neox.layers.17.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
+		<layer id="1789" name="__module.model.gpt_neox.layers.10.attention/aten::neg/Multiply" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2866,key.103">
+				<port id="2" precision="FP32" names="1543">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4251" name="__module.model.gpt_neox.layers.17.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4252" name="__module.model.gpt_neox.layers.17.attention/aten::size/Constant_2" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="4253" name="__module.model.gpt_neox.layers.17.attention/aten::size/Gather_2" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>10</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="2874" />
-			</output>
-		</layer>
-		<layer id="4254" name="Constant_337871_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="4255" name="Constant_337871" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
 			</output>
 		</layer>
-		<layer id="4256" name="Constant_337864" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
+		<layer id="1790" name="Constant_7125" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4257" name="Constant_337866" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
+		<layer id="1791" name="Constant_7169" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4258" name="Constant_337868" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
+		<layer id="1792" name="Constant_7171" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4259" name="Concat_337869" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="1793" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Slice_3" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -65061,163 +26937,125 @@
 				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
 				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4260" name="Broadcast_337872" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="5" precision="FP32" names="1537">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4261" name="ReadValue_330461" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.17.keypresent.17.key" variable_type="f32" variable_shape="?,32,?,80" />
+		<layer id="1794" name="__module.model.gpt_neox.layers.10.attention/aten::cat/Concat" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="past_key_values.17.key">
+				<port id="2" precision="FP32" names="1545">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4262" name="Constant_323487" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64" />
-			</output>
-		</layer>
-		<layer id="4263" name="Gather_323488" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="1795" name="__module.model.gpt_neox.layers.10.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="I32">
+				<port id="1" precision="FP32">
 					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
-				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="3" precision="FP32">
+				<port id="2" precision="FP32" names="1546">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4264" name="__module.model.gpt_neox.layers.17.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="1796" name="__module.model.gpt_neox.layers.10.attention/aten::add/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="1547">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4265" name="__module.model.gpt_neox.layers.17.attention/aten::size/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1797" name="Constant_7080" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="4266" name="__module.model.gpt_neox.layers.17.attention/aten::size/Gather_3" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="2876" />
-			</output>
-		</layer>
-		<layer id="4267" name="__module.model.gpt_neox.layers.17.attention/aten::add_/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" names="2875,2879,2880,seq_len.69" />
-			</output>
-		</layer>
-		<layer id="4268" name="__module.model.gpt_neox.layers.17.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
 			</output>
 		</layer>
-		<layer id="4269" name="__module.model.gpt_neox.layers.17.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="1798" name="Constant_7082" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4270" name="__module.model.gpt_neox.layers.17.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1799" name="Constant_7084" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4271" name="__module.model.gpt_neox.layers.17.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1800" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Slice_4" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -65228,197 +27066,192 @@
 				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2883,2886,cos.69">
+				<port id="5" precision="FP32" names="1527">
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4272" name="Constant_38299" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="4273" name="__module.model.gpt_neox.layers.17.attention/aten::index/Gather" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="1801" name="__module.model.gpt_neox.layers.10.attention/aten::cat/Concat_1" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
-				<port id="1" precision="I32">
+				<port id="1" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="2889">
+				<port id="2" precision="FP32" names="1565">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4274" name="__module.model.gpt_neox.layers.17.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64" />
-			</input>
+		<layer id="1802" name="Constant_52791" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="2" precision="FP32" names="2890,cos.71">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
+				<port id="0" precision="FP32" />
 			</output>
 		</layer>
-		<layer id="4275" name="__module.model.gpt_neox.layers.17.attention/aten::mul/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
+		<layer id="1803" name="Constant_52784" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
-				<port id="2" precision="FP32" names="2894">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4276" name="Constant_275439" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1804" name="Constant_52786" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4277" name="Constant_275438" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1805" name="Constant_52788" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4278" name="__module.model.gpt_neox.layers.17.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="1806" name="Concat_52789" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
+				<port id="4" precision="I64">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4279" name="__module.model.gpt_neox.layers.17.attention/aten::size/Constant_5" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="4280" name="__module.model.gpt_neox.layers.17.attention/aten::size/Gather_5" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="1807" name="Broadcast_52792" type="Broadcast" version="opset3">
+			<data mode="numpy" />
 			<input>
-				<port id="0" precision="I64">
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="2895,2900" />
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4281" name="__module.model.gpt_neox.layers.17.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
+		<layer id="1808" name="ReadValue_50191" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.10.keypresent.10.key" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64" />
+				<port id="1" precision="FP32" names="past_key_values.10.key">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4282" name="__module.model.gpt_neox.layers.17.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1">
+		<layer id="1809" name="Constant_47838" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64" />
+			</output>
+		</layer>
+		<layer id="1810" name="Gather_47839" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
-				<port id="0" precision="I64" />
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I32">
+					<dim>-1</dim>
+				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="1" precision="I64" names="2897,2898,2902,2903" />
+				<port id="3" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4283" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Constant_1" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1811" name="Constant_7034" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4284" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="1812" name="Constant_7036" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4285" name="Constant_275437" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1813" name="Constant_7038" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4286" name="ScatterUpdate_275440" type="ScatterUpdate" version="opset3">
+		<layer id="1814" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Slice_5" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -65426,148 +27259,155 @@
 				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
 				<port id="4" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="4287" name="Constant_275443" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="5" precision="FP32" names="1520">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4288" name="Constant_275446" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1815" name="Constant_7047" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="I64" names="1521">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4289" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1816" name="__module.model.gpt_neox.layers.10.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2904,x2.69">
+				<port id="2" precision="FP32" names="1522,key.41">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4290" name="Constant_299246_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
+		<layer id="1817" name="Constant_7090" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1818" name="Constant_7092" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1819" name="Constant_7094" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4291" name="Constant_299246" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="1820" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Slice_6" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="5" precision="FP32" names="1528,k.21">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4292" name="__module.model.gpt_neox.layers.17.attention/aten::neg/Multiply" type="Multiply" version="opset1">
+		<layer id="1821" name="__module.model.gpt_neox.layers.10.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
+					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2905">
+				<port id="2" precision="FP32" names="1548">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4293" name="Constant_275491" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1822" name="Constant_7371" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4294" name="Constant_275496" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1823" name="Constant_7373" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1824" name="Constant_7375" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4295" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1825" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Slice_7" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -65576,17 +27416,20 @@
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2899">
+				<port id="5" precision="FP32" names="1558,x2.43">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -65594,8 +27437,19 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4296" name="__module.model.gpt_neox.layers.17.attention/aten::cat/Concat" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="1826" name="Constant_37374" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1827" name="__module.model.gpt_neox.layers.10.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -65604,75 +27458,51 @@
 					<dim>10</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2907">
+				<port id="2" precision="FP32" names="1559">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4297" name="self.model.gpt_neox.layers.17.attention.rotary_emb.sin_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201162356" size="81920" />
-			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.17.attention.rotary_emb.sin_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4298" name="self.model.gpt_neox.layers.17.attention.rotary_emb.sin_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</input>
+		<layer id="1828" name="Constant_7268" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4299" name="__module.model.gpt_neox.layers.17.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1829" name="Constant_7312" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4300" name="__module.model.gpt_neox.layers.17.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1830" name="Constant_7314" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4301" name="__module.model.gpt_neox.layers.17.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1831" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Slice_8" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>2048</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
@@ -65684,60 +27514,45 @@
 				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2884,2887,sin.69">
+				<port id="5" precision="FP32" names="1553">
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4302" name="Constant_38302" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="4303" name="__module.model.gpt_neox.layers.17.attention/aten::index/Gather_1" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="1832" name="__module.model.gpt_neox.layers.10.attention/aten::cat/Concat_2" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="2892">
-					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>10</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="4304" name="__module.model.gpt_neox.layers.17.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
+				<port id="1" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>10</dim>
 				</port>
-				<port id="1" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2893,sin.71">
+				<port id="2" precision="FP32" names="1561">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4305" name="__module.model.gpt_neox.layers.17.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
+		<layer id="1833" name="__module.model.gpt_neox.layers.10.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -65754,7 +27569,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2908">
+				<port id="2" precision="FP32" names="1562">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -65762,7 +27577,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4306" name="__module.model.gpt_neox.layers.17.attention/aten::add/Add" type="Add" version="opset1">
+		<layer id="1834" name="__module.model.gpt_neox.layers.10.attention/aten::add/Add_1" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -65779,7 +27594,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2909">
+				<port id="2" precision="FP32" names="1563">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -65787,41 +27602,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4307" name="Constant_275542" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1835" name="Constant_7100" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4308" name="Constant_275545" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1836" name="Constant_7102" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4309" name="Constant_275548" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1837" name="Constant_7104" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4310" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1838" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Slice_9" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -65830,17 +27635,20 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2871">
+				<port id="5" precision="FP32" names="1529">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -65848,7 +27656,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4311" name="__module.model.gpt_neox.layers.17.attention/aten::cat/Concat_1" type="Concat" version="opset1">
+		<layer id="1839" name="__module.model.gpt_neox.layers.10.attention/aten::cat/Concat_3" type="Concat" version="opset1">
 			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
@@ -65865,7 +27673,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2927,2937,query.105,query.107">
+				<port id="2" precision="FP32" names="1567">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -65873,41 +27681,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4312" name="Constant_275554" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4313" name="Constant_275557" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4314" name="Constant_275560" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4315" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1840" name="__module.model.gpt_neox.layers.10.attention/aten::cat/Concat_4" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -65915,163 +27690,57 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="2872,k.35">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4316" name="__module.model.gpt_neox.layers.17.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2910">
+				<port id="2" precision="FP32" names="present.10.key">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4317" name="Constant_275564" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1841" name="Constant_52803" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
+				<port id="0" precision="FP32" />
 			</output>
 		</layer>
-		<layer id="4318" name="Constant_275563" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1842" name="Constant_52796" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4319" name="__module.model.gpt_neox.layers.17.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4320" name="__module.model.gpt_neox.layers.17.attention/aten::size/Constant_7" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1843" name="Constant_52798" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="4321" name="__module.model.gpt_neox.layers.17.attention/aten::size/Gather_7" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="2911,2916" />
-			</output>
-		</layer>
-		<layer id="4322" name="__module.model.gpt_neox.layers.17.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" />
-			</output>
-		</layer>
-		<layer id="4323" name="__module.model.gpt_neox.layers.17.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-			</input>
-			<output>
-				<port id="1" precision="I64" names="2913,2914,2918,2919" />
-			</output>
-		</layer>
-		<layer id="4324" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="4325" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4326" name="Constant_275562" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1844" name="Constant_52800" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4327" name="ScatterUpdate_275565" type="ScatterUpdate" version="opset3">
+		<layer id="1845" name="Concat_52801" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -66079,7 +27748,7 @@
 				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
@@ -66089,318 +27758,318 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4328" name="Constant_275568" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
+		<layer id="1846" name="Broadcast_52804" type="Broadcast" version="opset3">
+			<data mode="numpy" />
+			<input>
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="4329" name="Constant_275571" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4330" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1847" name="ReadValue_50193" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.10.valuepresent.10.value" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2920,x2.71">
+				<port id="1" precision="FP32" names="past_key_values.10.value">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4331" name="Constant_299247_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4332" name="Constant_299247" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
+		<layer id="1848" name="Constant_47841" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="4333" name="__module.model.gpt_neox.layers.17.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1849" name="Gather_47842" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="1" precision="I32">
+					<dim>-1</dim>
 				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2921">
+				<port id="3" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4334" name="Constant_275616" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1850" name="Constant_7052" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4335" name="Constant_275621" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1851" name="Constant_7054" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1852" name="Constant_7056" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4336" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1853" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Slice_10" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2915">
+				<port id="5" precision="FP32" names="1523">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4337" name="__module.model.gpt_neox.layers.17.attention/aten::cat/Concat_2" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="1854" name="Constant_7065" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
+			<output>
+				<port id="0" precision="I64" names="1524">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1855" name="__module.model.gpt_neox.layers.10.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2923">
+				<port id="2" precision="FP32" names="1525">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4338" name="__module.model.gpt_neox.layers.17.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1856" name="__module.model.gpt_neox.layers.10.attention/aten::cat/Concat_5" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2924">
+				<port id="2" precision="FP32" names="present.10.value">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4339" name="__module.model.gpt_neox.layers.17.attention/aten::add/Add_1" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1857" name="Constant_7447" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1858" name="ShapeOf_39422" type="ShapeOf" version="opset3">
+			<data output_type="i64" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2925">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4340" name="Constant_275666" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1859" name="Constant_39423" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116168" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4341" name="Constant_275669" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1860" name="Constant_39424" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
+				<port id="0" precision="I64" />
+			</output>
+		</layer>
+		<layer id="1861" name="Gather_39425" type="Gather" version="opset8">
+			<data batch_dims="0" />
+			<input>
 				<port id="0" precision="I64">
 					<dim>4</dim>
 				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64" />
+			</input>
+			<output>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4342" name="Constant_275672" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1862" name="Add_39428" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="I64" names="1572">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1863" name="Constant_7450" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4343" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1864" name="__module.model.gpt_neox.layers.10.attention/aten::slice/Slice_14" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2873">
+				<port id="5" precision="FP32" names="1576">
+					<dim>-1</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>60</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4344" name="__module.model.gpt_neox.layers.17.attention/aten::cat/Concat_3" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="1865" name="__module.model.gpt_neox.layers.10.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+			<data causal="false" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>60</dim>
+					<dim>80</dim>
+				</port>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="3" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2929">
+				<port id="4" precision="FP32" names="1577,attn_output.61">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -66408,8 +28077,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4345" name="__module.model.gpt_neox.layers.17.attention/aten::to/Convert_1" type="Concat" version="opset1">
-			<data axis="-2" />
+		<layer id="1866" name="__module.model.gpt_neox.layers.10.attention/aten::transpose/Constant" type="Const" version="opset1">
+			<data element_type="i32" shape="4" offset="201116200" size="16" />
+			<output>
+				<port id="0" precision="I32">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1867" name="__module.model.gpt_neox.layers.10.attention/aten::transpose/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -66417,869 +28093,1044 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
+				<port id="1" precision="I32">
+					<dim>4</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="1578">
+					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
+					<dim>80</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1868" name="Constant_40994" type="Const" version="opset1">
+			<data element_type="i64" shape="3" offset="201116216" size="24" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>3</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1869" name="__module.model.gpt_neox.layers.10.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
+			<data special_zero="true" />
+			<input>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>80</dim>
 				</port>
+				<port id="1" precision="I64">
+					<dim>3</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="present.17.key">
+				<port id="2" precision="FP32" names="1581">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4346" name="Constant_337884_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
+		<layer id="1870" name="self.model.gpt_neox.layers.10.attention.dense.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 2560" offset="989314640" size="6553600" />
 			<output>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4347" name="Constant_337884" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="1871" name="Convert_75286" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" />
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4348" name="Constant_337877" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
+		<layer id="1872" name="self.model.gpt_neox.layers.10.attention.dense.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="995868240" size="2560" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="U8">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4349" name="Constant_337879" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64">
+		<layer id="1873" name="Convert_75289" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="4350" name="Constant_337881" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
+			</input>
 			<output>
-				<port id="0" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4351" name="Concat_337882" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="1874" name="self.model.gpt_neox.layers.10.attention.dense.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1875" name="self.model.gpt_neox.layers.10.attention.dense.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="995870800" size="5120" />
+			<output>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I64">
+			</output>
+		</layer>
+		<layer id="1876" name="self.model.gpt_neox.layers.10.attention.dense.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4352" name="Broadcast_337885" type="Broadcast" version="opset3">
-			<data mode="numpy" />
+		<layer id="1877" name="self.model.gpt_neox.layers.10.attention.dense.weight/fq_weights_1/convert" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4353" name="ReadValue_330463" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.17.valuepresent.17.value" variable_type="f32" variable_shape="?,32,?,80" />
+		<layer id="1878" name="__module.model.gpt_neox.layers.10.attention.dense/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="past_key_values.17.value">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4354" name="Constant_323490" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
+		<layer id="1879" name="Constant_37375" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="995875920" size="10240" />
 			<output>
-				<port id="0" precision="I64" />
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4355" name="Gather_323491" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="1880" name="__module.model.gpt_neox.layers.10.attention.dense/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="3" precision="FP32">
+				<port id="2" precision="FP32" names="1584,1586,input.41">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4356" name="Constant_275678" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+		<layer id="1881" name="__module.model.gpt_neox.layers.10/aten::add/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="4357" name="Constant_275681" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="4358" name="Constant_275684" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="1604">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4359" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1882" name="__module.model.gpt_neox.layers.10/aten::add/Add_1" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2867">
+				<port id="2" precision="FP32" names="1605,1607,hidden_states.47">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4360" name="Constant_10128" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="1883" name="__module.model.gpt_neox.layers.11.input_layernorm/aten::layer_norm/Multiply" type="Const" version="opset1">
+			<data element_type="i32" shape="1" offset="128867668" size="4" />
 			<output>
-				<port id="0" precision="I64" names="2868">
-					<dim>4</dim>
+				<port id="0" precision="I32">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4361" name="__module.model.gpt_neox.layers.17.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
+		<layer id="1884" name="__module.model.gpt_neox.layers.11.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
+			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="I32">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2869">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4362" name="__module.model.gpt_neox.layers.17.attention/aten::cat/Concat_5" type="Concat" version="opset1">
-			<data axis="-2" />
+		<layer id="1885" name="Constant_37376" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="995886160" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1886" name="__module.model.gpt_neox.layers.11.post_attention_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="present.17.value">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4363" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Slice_11" type="Const" version="opset1">
-			<data element_type="boolean" shape="1, 1, 2048, 2048" offset="201244276" size="4194304" />
+		<layer id="1887" name="Constant_37377" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="995896400" size="10240" />
 			<output>
-				<port id="0" precision="BOOL" names="2946">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4364" name="Constant_275688" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1888" name="__module.model.gpt_neox.layers.11.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="2" precision="FP32" names="1705">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4365" name="Constant_275687" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162298" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1889" name="self.model.gpt_neox.layers.11.mlp.dense_h_to_4h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 2560" offset="995906640" size="26214400" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4366" name="__module.model.gpt_neox.layers.17.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="1890" name="Convert_74404" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4367" name="__module.model.gpt_neox.layers.17.attention/aten::size/Constant_8" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1891" name="self.model.gpt_neox.layers.11.mlp.dense_h_to_4h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 1" offset="1022121040" size="10240" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4368" name="__module.model.gpt_neox.layers.17.attention/aten::size/Gather_8" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="1892" name="Convert_74407" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="2941" />
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4369" name="__module.model.gpt_neox.layers.17.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="1893" name="self.model.gpt_neox.layers.11.mlp.dense_h_to_4h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4370" name="__module.model.gpt_neox.layers.17.attention/aten::size/Constant_9" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1894" name="self.model.gpt_neox.layers.11.mlp.dense_h_to_4h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="10240, 1" offset="1022131280" size="20480" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="4371" name="__module.model.gpt_neox.layers.17.attention/aten::size/Gather_9" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="2939" />
 			</output>
 		</layer>
-		<layer id="4372" name="__module.model.gpt_neox.layers.17.attention/aten::sub/Subtract" type="Subtract" version="opset1">
+		<layer id="1895" name="self.model.gpt_neox.layers.11.mlp.dense_h_to_4h.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64" names="2943,2944" />
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4373" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Constant_4" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
+		<layer id="1896" name="__module.model.gpt_neox.layers.11.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<rt_info>
-				<attribute name="precise" version="0" />
+				<attribute name="decompression" version="0" />
 			</rt_info>
+			<input>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I32" />
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4374" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1">
+		<layer id="1897" name="__module.model.gpt_neox.layers.11.mlp.dense_h_to_4h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1898" name="self.model.gpt_neox.layers.11.mlp.dense_h_to_4h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="10240" offset="1022151760" size="20480" />
+			<output>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.11.mlp.dense_h_to_4h.bias">
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4375" name="Constant_275686" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
+		<layer id="1899" name="__module.model.gpt_neox.layers.11.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<rt_info>
-				<attribute name="precise" version="0" />
+				<attribute name="decompression" version="0" />
 			</rt_info>
+			<input>
+				<port id="0" precision="BF16">
+					<dim>10240</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4376" name="ScatterUpdate_275689" type="ScatterUpdate" version="opset3">
+		<layer id="1900" name="__module.model.gpt_neox.layers.11.mlp.dense_h_to_4h/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
+				<port id="2" precision="FP32" names="1710">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4377" name="Constant_275690" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1901" name="__module.model.gpt_neox.layers.11.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
+			<data approximation_mode="ERF" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="1" precision="FP32" names="1711">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4378" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="1902" name="self.model.gpt_neox.layers.11.mlp.dense_4h_to_h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 10240" offset="1022172240" size="26214400" />
 			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4379" name="ScatterUpdate_275691" type="ScatterUpdate" version="opset3">
+		<layer id="1903" name="Convert_74414" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4380" name="Constant_275694" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438612" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1904" name="self.model.gpt_neox.layers.11.mlp.dense_4h_to_h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="1048386640" size="2560" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4381" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1905" name="Convert_74417" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
 					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>3</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="BOOL" names="2947">
-					<dim>1</dim>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4382" name="Constant_275767" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+		<layer id="1906" name="self.model.gpt_neox.layers.11.mlp.dense_4h_to_h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="4383" name="Constant_275768" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="4384" name="Constant_275764" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4385" name="Constant_275763" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1907" name="self.model.gpt_neox.layers.11.mlp.dense_4h_to_h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="1048389200" size="5120" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4386" name="ScatterUpdate_275769" type="ScatterUpdate" version="opset3">
+		<layer id="1908" name="self.model.gpt_neox.layers.11.mlp.dense_4h_to_h.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4387" name="Constant_275772" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
+		<layer id="1909" name="__module.model.gpt_neox.layers.11.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<rt_info>
-				<attribute name="precise" version="0" />
+				<attribute name="decompression" version="0" />
 			</rt_info>
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4388" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="1910" name="__module.model.gpt_neox.layers.11.mlp.dense_4h_to_h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="BOOL" names="2948,2949">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4389" name="__module.model.gpt_neox.layers.17.attention/aten::where/Convert_1_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
+		<layer id="1911" name="self.model.gpt_neox.layers.11.mlp.dense_4h_to_h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="2560" offset="1048394320" size="5120" />
 			<output>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.11.mlp.dense_4h_to_h.bias">
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4390" name="__module.model.gpt_neox.layers.17.attention/aten::where/Convert_1" type="Convert" version="opset1">
+		<layer id="1912" name="__module.model.gpt_neox.layers.11.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
 			</rt_info>
 			<input>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="BF16">
+					<dim>2560</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="4391" name="__module.model.gpt_neox.layers.17.attention/aten::full/Convert" type="Const" version="opset1">
-			<data element_type="f32" shape="" offset="205438636" size="4" />
-			<output>
-				<port id="0" precision="FP32" names="2936" />
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4392" name="__module.model.gpt_neox.layers.17.attention/aten::where/Select" type="Select" version="opset1">
+		<layer id="1913" name="__module.model.gpt_neox.layers.11.mlp.dense_4h_to_h/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="FP32" />
-				<port id="2" precision="FP32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="2950,causal_mask.69">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="1714,input.47">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4393" name="Add_290440" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
+		<layer id="1914" name="Constant_37378" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1048399440" size="10240" />
+			<output>
 				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1915" name="__module.model.gpt_neox.layers.11.input_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4394" name="ShapeOf_290441" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="1916" name="Constant_37379" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1048409680" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1917" name="__module.model.gpt_neox.layers.11.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
 					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="1617,hidden_states.45">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1918" name="self.model.gpt_neox.layers.11.attention.query_key_value.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 2560" offset="1048419920" size="19660800" />
+			<output>
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1919" name="Convert_74945" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4395" name="Constant_254584" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1920" name="self.model.gpt_neox.layers.11.attention.query_key_value.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 1" offset="1068080720" size="7680" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="U8">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4396" name="__module.model.gpt_neox.layers.17.attention/aten::size/Constant_10" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1921" name="Convert_74948" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>1</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I32" />
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4397" name="__module.model.gpt_neox.layers.17.attention/aten::size/Gather_10" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="1922" name="self.model.gpt_neox.layers.11.attention.query_key_value.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="2934">
+				<port id="2" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1923" name="self.model.gpt_neox.layers.11.attention.query_key_value.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="7680, 1" offset="1068088400" size="15360" />
+			<output>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4398" name="__module.model.gpt_neox.layers.17.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="1924" name="self.model.gpt_neox.layers.11.attention.query_key_value.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
+			</output>
+		</layer>
+		<layer id="1925" name="self.model.gpt_neox.layers.11.attention.query_key_value.weight/fq_weights_1/convert" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64" names="2951">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4399" name="__module.model.gpt_neox.layers.17.attention/aten::expand/Abs" type="Abs" version="opset1">
+		<layer id="1926" name="__module.model.gpt_neox.layers.11.attention.query_key_value/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>7680</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1927" name="Constant_37380" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 7680" offset="1068103760" size="30720" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4400" name="Maximum_290442" type="Maximum" version="opset1">
+		<layer id="1928" name="__module.model.gpt_neox.layers.11.attention.query_key_value/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>7680</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>7680</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
+				<port id="2" precision="FP32" names="1624,qkv.45">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>7680</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1929" name="__module.model.gpt_neox.layers.11.attention/prim::ListConstruct" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201115992" size="32" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4401" name="__module.model.gpt_neox.layers.17.attention/aten::add/Add_2" type="Broadcast" version="opset3">
-			<data mode="numpy" />
+		<layer id="1930" name="__module.model.gpt_neox.layers.11.attention/aten::view/Reshape" type="Reshape" version="opset1">
+			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>7680</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2953">
-					<dim>-1</dim>
-					<dim>-1</dim>
+				<port id="2" precision="FP32" names="1628,qkv.47">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4402" name="__module.model.gpt_neox.layers.17.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
-			<data causal="false" />
+		<layer id="1931" name="Constant_7646" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1932" name="Constant_7648" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1933" name="Constant_7650" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1934" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Slice" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>240</dim>
 				</port>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+				<port id="1" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="2954,2955,sdpa_result.35,tensor.69">
+				<port id="5" precision="FP32" names="1629">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4403" name="Constant_10569" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="1935" name="Constant_7659" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64" names="2956">
+				<port id="0" precision="I64" names="1630">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4404" name="__module.model.gpt_neox.layers.17.attention/aten::permute/Transpose_3" type="Transpose" version="opset1">
+		<layer id="1936" name="__module.model.gpt_neox.layers.11.attention/aten::permute/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
@@ -67287,1154 +29138,1135 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2957">
-					<dim>-1</dim>
+				<port id="2" precision="FP32" names="1631,query.23">
 					<dim>-1</dim>
 					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4405" name="__module.model.gpt_neox.layers.17.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438650" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="1937" name="Constant_7700" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>3</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4406" name="__module.model.gpt_neox.layers.17.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
-			<data special_zero="true" />
+		<layer id="1938" name="Constant_7702" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1939" name="Constant_7704" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1940" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Slice_1" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>3</dim>
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2962">
+				<port id="5" precision="FP32" names="1638,q.23">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4407" name="self.model.gpt_neox.layers.17.attention.dense.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 2560" offset="1544679634" size="6553600" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4408" name="Convert_22114" type="Convert" version="opset1">
-			<data destination_type="f16" />
+		<layer id="1941" name="__module.model.gpt_neox.layers.11.attention/aten::mul/Multiply" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32" names="1644">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4409" name="self.model.gpt_neox.layers.17.attention.dense.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="1551233234" size="2560" />
+		<layer id="1942" name="Constant_7858" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4410" name="Convert_22117" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
+		<layer id="1943" name="Constant_7860" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="1944" name="Constant_7862" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4411" name="self.model.gpt_neox.layers.17.attention.dense.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1945" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Slice_2" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4412" name="self.model.gpt_neox.layers.17.attention.dense.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="1551235794" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
+				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="4413" name="self.model.gpt_neox.layers.17.attention.dense.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="3" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="5" precision="FP32" names="1654,x2.45">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4414" name="self.model.gpt_neox.layers.17.attention.dense.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="1946" name="Constant_37381" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4415" name="__module.model.gpt_neox.layers.17.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
+		<layer id="1947" name="__module.model.gpt_neox.layers.11.attention/aten::neg/Multiply" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>2560</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="1655">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4416" name="Constant_299248_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1551240914" size="5120" />
+		<layer id="1948" name="Constant_7755" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4417" name="Constant_299248" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+		<layer id="1949" name="Constant_7799" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="1950" name="Constant_7801" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4418" name="__module.model.gpt_neox.layers.17.attention.dense/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1951" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Slice_3" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="FP32">
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2965,2967,input.71">
+				<port id="5" precision="FP32" names="1649">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4419" name="__module.model.gpt_neox.layers.17/aten::add/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1952" name="__module.model.gpt_neox.layers.11.attention/aten::cat/Concat" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2985">
+				<port id="2" precision="FP32" names="1657">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4420" name="__module.model.gpt_neox.layers.17/aten::add/Add_1" type="Add" version="opset1">
+		<layer id="1953" name="__module.model.gpt_neox.layers.11.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2986,2988,hidden_states.37">
+				<port id="2" precision="FP32" names="1658">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4421" name="__module.model.gpt_neox.layers.18.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867670" size="4" />
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4422" name="__module.model.gpt_neox.layers.18.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
-			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
+		<layer id="1954" name="__module.model.gpt_neox.layers.11.attention/aten::add/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="I32">
-					<dim>1</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="1659">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4423" name="Constant_299249_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1551246034" size="5120" />
+		<layer id="1955" name="Constant_7710" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4424" name="Constant_299249" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+		<layer id="1956" name="Constant_7712" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="1957" name="Constant_7714" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4425" name="__module.model.gpt_neox.layers.18.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1958" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Slice_4" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="1" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="4426" name="Constant_299250_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1551251154" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="2" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="4427" name="Constant_299250" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
+				<port id="3" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+				<port id="5" precision="FP32" names="1639">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4428" name="__module.model.gpt_neox.layers.18.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1959" name="__module.model.gpt_neox.layers.11.attention/aten::cat/Concat_1" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3123">
+				<port id="2" precision="FP32" names="1677">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4429" name="self.model.gpt_neox.layers.18.mlp.dense_h_to_4h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 2560" offset="1551256274" size="26214400" />
+		<layer id="1960" name="Constant_52815" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
+				<port id="0" precision="FP32" />
 			</output>
 		</layer>
-		<layer id="4430" name="Convert_22164" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="1961" name="Constant_52808" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4431" name="self.model.gpt_neox.layers.18.mlp.dense_h_to_4h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 1" offset="1577470674" size="10240" />
+		<layer id="1962" name="Constant_52810" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4432" name="Convert_22167" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</input>
+		<layer id="1963" name="Constant_52812" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4433" name="self.model.gpt_neox.layers.18.mlp.dense_h_to_4h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1964" name="Concat_52813" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
+				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="2" precision="I64">
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="4434" name="self.model.gpt_neox.layers.18.mlp.dense_h_to_4h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="10240, 1" offset="1577480914" size="20480" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
+			</input>
+			<output>
+				<port id="4" precision="I64">
+					<dim>4</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4435" name="self.model.gpt_neox.layers.18.mlp.dense_h_to_4h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1965" name="Broadcast_52816" type="Broadcast" version="opset3">
+			<data mode="numpy" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4436" name="self.model.gpt_neox.layers.18.mlp.dense_h_to_4h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="1966" name="ReadValue_50195" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.11.keypresent.11.key" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="1" precision="FP32" names="past_key_values.11.key">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4437" name="__module.model.gpt_neox.layers.18.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
+		<layer id="1967" name="Constant_47844" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64" />
+			</output>
+		</layer>
+		<layer id="1968" name="Gather_47845" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="1" precision="I32">
+					<dim>-1</dim>
 				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="3" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4438" name="Constant_299251_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 10240" offset="1577501394" size="20480" />
+		<layer id="1969" name="Constant_7664" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4439" name="Constant_299251" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+		<layer id="1970" name="Constant_7666" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>10240</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="1971" name="Constant_7668" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4440" name="__module.model.gpt_neox.layers.18.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1972" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Slice_5" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
-				<port id="1" precision="FP32">
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3128">
+				<port id="5" precision="FP32" names="1632">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4441" name="__module.model.gpt_neox.layers.18.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
-			<data approximation_mode="ERF" />
+		<layer id="1973" name="Constant_7677" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
+			<output>
+				<port id="0" precision="I64" names="1633">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1974" name="__module.model.gpt_neox.layers.11.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="3129">
+				<port id="2" precision="FP32" names="1634,key.45">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4442" name="self.model.gpt_neox.layers.18.mlp.dense_4h_to_h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 10240" offset="1577521874" size="26214400" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4443" name="Convert_22174" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
+		<layer id="1975" name="Constant_7720" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4444" name="self.model.gpt_neox.layers.18.mlp.dense_4h_to_h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="1603736274" size="2560" />
+		<layer id="1976" name="Constant_7722" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4445" name="Convert_22177" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
+		<layer id="1977" name="Constant_7724" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4446" name="self.model.gpt_neox.layers.18.mlp.dense_4h_to_h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1978" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Slice_6" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4447" name="self.model.gpt_neox.layers.18.mlp.dense_4h_to_h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="1603738834" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
+				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="4448" name="self.model.gpt_neox.layers.18.mlp.dense_4h_to_h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4449" name="self.model.gpt_neox.layers.18.mlp.dense_4h_to_h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="5" precision="FP32" names="1640,k.23">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4450" name="__module.model.gpt_neox.layers.18.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
+		<layer id="1979" name="__module.model.gpt_neox.layers.11.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="1660">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4451" name="Constant_299252_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1603743954" size="5120" />
+		<layer id="1980" name="Constant_8001" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4452" name="Constant_299252" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+		<layer id="1981" name="Constant_8003" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="1982" name="Constant_8005" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4453" name="__module.model.gpt_neox.layers.18.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1983" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Slice_7" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="FP32">
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3132,input.77">
+				<port id="5" precision="FP32" names="1670,x2.47">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4454" name="Constant_299253_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1603749074" size="5120" />
+		<layer id="1984" name="Constant_37382" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4455" name="Constant_299253" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4456" name="__module.model.gpt_neox.layers.18.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
+		<layer id="1985" name="__module.model.gpt_neox.layers.11.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="1671">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4457" name="Constant_299254_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1603754194" size="5120" />
+		<layer id="1986" name="Constant_7898" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="FP16">
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1987" name="Constant_7942" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="1988" name="Constant_7944" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4458" name="Constant_299254" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="1989" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Slice_8" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
+				<port id="3" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="5" precision="FP32" names="1665">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4459" name="__module.model.gpt_neox.layers.18.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1990" name="__module.model.gpt_neox.layers.11.attention/aten::cat/Concat_2" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="2998">
+				<port id="2" precision="FP32" names="1673">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4460" name="self.model.gpt_neox.layers.18.attention.query_key_value.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 2560" offset="1603759314" size="19660800" />
+		<layer id="1991" name="__module.model.gpt_neox.layers.11.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32" names="1674">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4461" name="Convert_22144" type="Convert" version="opset1">
-			<data destination_type="f16" />
+		<layer id="1992" name="__module.model.gpt_neox.layers.11.attention/aten::add/Add_1" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32" names="1675">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4462" name="self.model.gpt_neox.layers.18.attention.query_key_value.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 1" offset="1623420114" size="7680" />
+		<layer id="1993" name="Constant_7730" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4463" name="Convert_22147" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
+		<layer id="1994" name="Constant_7732" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="1995" name="Constant_7734" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4464" name="self.model.gpt_neox.layers.18.attention.query_key_value.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="1996" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Slice_9" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
+				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4465" name="self.model.gpt_neox.layers.18.attention.query_key_value.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="7680, 1" offset="1623427794" size="15360" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
+				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="4466" name="self.model.gpt_neox.layers.18.attention.query_key_value.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="3" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="5" precision="FP32" names="1641">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4467" name="self.model.gpt_neox.layers.18.attention.query_key_value.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="1997" name="__module.model.gpt_neox.layers.11.attention/aten::cat/Concat_3" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32" names="1679">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4468" name="__module.model.gpt_neox.layers.18.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
+		<layer id="1998" name="__module.model.gpt_neox.layers.11.attention/aten::cat/Concat_4" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>7680</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="present.11.key">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4469" name="Constant_299255_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 7680" offset="1623443154" size="15360" />
+		<layer id="1999" name="Constant_52827" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32" />
+			</output>
+		</layer>
+		<layer id="2000" name="Constant_52820" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2001" name="Constant_52822" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2002" name="Constant_52824" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4470" name="Constant_299255" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="2003" name="Concat_52825" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>1</dim>
-					<dim>7680</dim>
 				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
+				<port id="2" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
 					<dim>1</dim>
-					<dim>7680</dim>
+				</port>
+			</input>
+			<output>
+				<port id="4" precision="I64">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4471" name="__module.model.gpt_neox.layers.18.attention.query_key_value/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2004" name="Broadcast_52828" type="Broadcast" version="opset3">
+			<data mode="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
+					<dim>4</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
+			</output>
+		</layer>
+		<layer id="2005" name="ReadValue_50197" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.11.valuepresent.11.value" variable_type="f32" variable_shape="?,32,?,80" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3005,qkv.73">
+				<port id="1" precision="FP32" names="past_key_values.11.value">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4472" name="__module.model.gpt_neox.layers.18.attention/prim::ListConstruct/Concat" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080154" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2006" name="Constant_47847" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="4473" name="__module.model.gpt_neox.layers.18.attention/aten::view/Reshape" type="Reshape" version="opset1">
-			<data special_zero="true" />
+		<layer id="2007" name="Gather_47848" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="I32">
+					<dim>-1</dim>
 				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3009,qkv.75">
-					<dim>-1</dim>
+				<port id="3" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>240</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4474" name="Constant_275779" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2008" name="Constant_7682" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4475" name="Constant_275782" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2009" name="Constant_7684" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4476" name="Constant_275785" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2010" name="Constant_7686" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4477" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2011" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Slice_10" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -68443,17 +30275,20 @@
 					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3010">
+				<port id="5" precision="FP32" names="1635">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -68461,15 +30296,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4478" name="Constant_10681" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="2012" name="Constant_7695" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64" names="3011">
+				<port id="0" precision="I64" names="1636">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4479" name="__module.model.gpt_neox.layers.18.attention/aten::permute/Transpose" type="Transpose" version="opset1">
+		<layer id="2013" name="__module.model.gpt_neox.layers.11.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -68482,7 +30317,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3012,query.109">
+				<port id="2" precision="FP32" names="1637">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -68490,41 +30325,41 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4480" name="Constant_275791" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+		<layer id="2014" name="__module.model.gpt_neox.layers.11.attention/aten::cat/Concat_5" type="Concat" version="opset1">
+			<data axis="-2" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="4481" name="Constant_275794" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="present.11.value">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4482" name="Constant_275797" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2015" name="Constant_8077" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4483" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2016" name="ShapeOf_39493" type="ShapeOf" version="opset3">
+			<data output_type="i64" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -68532,1053 +30367,1197 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
+			</input>
+			<output>
 				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="2017" name="Constant_39494" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116168" size="8" />
 			<output>
-				<port id="4" precision="FP32" names="3019,q.37">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4484" name="self.model.gpt_neox.layers.18.attention.rotary_emb.cos_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201080346" size="81920" />
+		<layer id="2018" name="Constant_39495" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.18.attention.rotary_emb.cos_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="4485" name="self.model.gpt_neox.layers.18.attention.rotary_emb.cos_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="2019" name="Gather_39496" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="I64">
+					<dim>4</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
 				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="3" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4486" name="__module.model.gpt_neox.layers.18.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
+		<layer id="2020" name="Add_39499" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="I64" names="1684">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4487" name="Constant_275803" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2021" name="Constant_8080" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4488" name="Constant_275806" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+		<layer id="2022" name="__module.model.gpt_neox.layers.11.attention/aten::slice/Slice_14" type="Slice" version="opset8">
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="4489" name="Constant_275809" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="5" precision="FP32" names="1688">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4490" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2023" name="__module.model.gpt_neox.layers.11.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+			<data causal="false" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>240</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="3" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3013">
-					<dim>-1</dim>
+				<port id="4" precision="FP32" names="1689,attn_output.67">
 					<dim>-1</dim>
 					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4491" name="Constant_10695" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="2024" name="__module.model.gpt_neox.layers.11.attention/aten::transpose/Constant" type="Const" version="opset1">
+			<data element_type="i32" shape="4" offset="201116200" size="16" />
 			<output>
-				<port id="0" precision="I64" names="3014">
+				<port id="0" precision="I32">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4492" name="__module.model.gpt_neox.layers.18.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
+		<layer id="2025" name="__module.model.gpt_neox.layers.11.attention/aten::transpose/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="I32">
 					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3015,key.109">
+				<port id="2" precision="FP32" names="1690">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4493" name="__module.model.gpt_neox.layers.18.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="2026" name="Constant_40995" type="Const" version="opset1">
+			<data element_type="i64" shape="3" offset="201116216" size="24" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>3</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2027" name="__module.model.gpt_neox.layers.11.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
+			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>80</dim>
 				</port>
+				<port id="1" precision="I64">
+					<dim>3</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="1693">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4494" name="__module.model.gpt_neox.layers.18.attention/aten::size/Constant_2" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2028" name="self.model.gpt_neox.layers.11.attention.dense.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 2560" offset="1068134480" size="6553600" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4495" name="__module.model.gpt_neox.layers.18.attention/aten::size/Gather_2" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="2029" name="Convert_75297" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="3023" />
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4496" name="Constant_337897_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
+		<layer id="2030" name="self.model.gpt_neox.layers.11.attention.dense.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="1074688080" size="2560" />
 			<output>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4497" name="Constant_337897" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="2031" name="Convert_75300" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="4498" name="Constant_337890" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
-			<output>
-				<port id="0" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4499" name="Constant_337892" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64">
+		<layer id="2032" name="self.model.gpt_neox.layers.11.attention.dense.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4500" name="Constant_337894" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
+		<layer id="2033" name="self.model.gpt_neox.layers.11.attention.dense.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="1074690640" size="5120" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4501" name="Concat_337895" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="2034" name="self.model.gpt_neox.layers.11.attention.dense.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
+			</output>
+		</layer>
+		<layer id="2035" name="self.model.gpt_neox.layers.11.attention.dense.weight/fq_weights_1/convert" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4502" name="Broadcast_337898" type="Broadcast" version="opset3">
-			<data mode="numpy" />
+		<layer id="2036" name="__module.model.gpt_neox.layers.11.attention.dense/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4503" name="ReadValue_330465" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.18.keypresent.18.key" variable_type="f32" variable_shape="?,32,?,80" />
+		<layer id="2037" name="Constant_37383" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1074695760" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2038" name="__module.model.gpt_neox.layers.11.attention.dense/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="past_key_values.18.key">
+				<port id="2" precision="FP32" names="1696,1698,input.45">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4504" name="Constant_323493" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
+		<layer id="2039" name="__module.model.gpt_neox.layers.11/aten::add/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64" />
+				<port id="2" precision="FP32" names="1716">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4505" name="Gather_323494" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="2040" name="__module.model.gpt_neox.layers.11/aten::add/Add_1" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I32">
+				<port id="1" precision="FP32">
 					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="3" precision="FP32">
+				<port id="2" precision="FP32" names="1717,1719,hidden_states.51">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4506" name="__module.model.gpt_neox.layers.18.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="2041" name="__module.model.gpt_neox.layers.12.input_layernorm/aten::layer_norm/Multiply" type="Const" version="opset1">
+			<data element_type="i32" shape="1" offset="128867668" size="4" />
+			<output>
+				<port id="0" precision="I32">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2042" name="__module.model.gpt_neox.layers.12.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
+			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="I32">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4507" name="__module.model.gpt_neox.layers.18.attention/aten::size/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2043" name="Constant_37384" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1074706000" size="10240" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4508" name="__module.model.gpt_neox.layers.18.attention/aten::size/Gather_3" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="2044" name="__module.model.gpt_neox.layers.12.post_attention_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="3025" />
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2045" name="Constant_37385" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1074716240" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4509" name="__module.model.gpt_neox.layers.18.attention/aten::add_/Add" type="Add" version="opset1">
+		<layer id="2046" name="__module.model.gpt_neox.layers.12.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64" names="3024,3028,3029,seq_len.73" />
+				<port id="2" precision="FP32" names="1817">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4510" name="__module.model.gpt_neox.layers.18.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2047" name="self.model.gpt_neox.layers.12.mlp.dense_h_to_4h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 2560" offset="1074726480" size="26214400" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4511" name="__module.model.gpt_neox.layers.18.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1">
+		<layer id="2048" name="Convert_74424" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4512" name="__module.model.gpt_neox.layers.18.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2049" name="self.model.gpt_neox.layers.12.mlp.dense_h_to_4h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 1" offset="1100940880" size="10240" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="U8">
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4513" name="__module.model.gpt_neox.layers.18.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2050" name="Convert_74427" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
+				<port id="0" precision="U8">
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
+			</input>
+			<output>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I64">
+			</output>
+		</layer>
+		<layer id="2051" name="self.model.gpt_neox.layers.12.mlp.dense_h_to_4h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3032,3035,cos.73">
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4514" name="Constant_38305" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
+		<layer id="2052" name="self.model.gpt_neox.layers.12.mlp.dense_h_to_4h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="10240, 1" offset="1100951120" size="20480" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4515" name="__module.model.gpt_neox.layers.18.attention/aten::index/Gather" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="2053" name="self.model.gpt_neox.layers.12.mlp.dense_h_to_4h.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
 				</port>
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="3038">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4516" name="__module.model.gpt_neox.layers.18.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1">
+		<layer id="2054" name="__module.model.gpt_neox.layers.12.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<rt_info>
+				<attribute name="decompression" version="0" />
+			</rt_info>
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3039,cos.75">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4517" name="__module.model.gpt_neox.layers.18.attention/aten::mul/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2055" name="__module.model.gpt_neox.layers.12.mlp.dense_h_to_4h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3043">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4518" name="Constant_275814" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2056" name="self.model.gpt_neox.layers.12.mlp.dense_h_to_4h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="10240" offset="1100971600" size="20480" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.12.mlp.dense_h_to_4h.bias">
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4519" name="Constant_275813" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
+		<layer id="2057" name="__module.model.gpt_neox.layers.12.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<rt_info>
-				<attribute name="precise" version="0" />
+				<attribute name="decompression" version="0" />
 			</rt_info>
+			<input>
+				<port id="0" precision="BF16">
+					<dim>10240</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4520" name="__module.model.gpt_neox.layers.18.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="2058" name="__module.model.gpt_neox.layers.12.mlp.dense_h_to_4h/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>10240</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="1822">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4521" name="__module.model.gpt_neox.layers.18.attention/aten::size/Constant_5" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="4522" name="__module.model.gpt_neox.layers.18.attention/aten::size/Gather_5" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="2059" name="__module.model.gpt_neox.layers.12.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
+			<data approximation_mode="ERF" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="3044,3049" />
+				<port id="1" precision="FP32" names="1823">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4523" name="__module.model.gpt_neox.layers.18.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
+		<layer id="2060" name="self.model.gpt_neox.layers.12.mlp.dense_4h_to_h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 10240" offset="1100992080" size="26214400" />
 			<output>
-				<port id="2" precision="I64" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4524" name="__module.model.gpt_neox.layers.18.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1">
+		<layer id="2061" name="Convert_74434" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64" names="3046,3047,3051,3052" />
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4525" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Constant_1" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2062" name="self.model.gpt_neox.layers.12.mlp.dense_4h_to_h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="1127206480" size="2560" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2063" name="Convert_74437" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4526" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1">
+		<layer id="2064" name="self.model.gpt_neox.layers.12.mlp.dense_4h_to_h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4527" name="Constant_275812" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2065" name="self.model.gpt_neox.layers.12.mlp.dense_4h_to_h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="1127209040" size="5120" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4528" name="ScatterUpdate_275815" type="ScatterUpdate" version="opset3">
+		<layer id="2066" name="self.model.gpt_neox.layers.12.mlp.dense_4h_to_h.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4529" name="Constant_275818" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
+		<layer id="2067" name="__module.model.gpt_neox.layers.12.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<rt_info>
-				<attribute name="precise" version="0" />
+				<attribute name="decompression" version="0" />
 			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="4530" name="Constant_275821" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4531" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2068" name="__module.model.gpt_neox.layers.12.mlp.dense_4h_to_h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3053,x2.73">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4532" name="Constant_299256_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
+		<layer id="2069" name="self.model.gpt_neox.layers.12.mlp.dense_4h_to_h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="2560" offset="1127214160" size="5120" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.12.mlp.dense_4h_to_h.bias">
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4533" name="Constant_299256" type="Convert" version="opset1">
+		<layer id="2070" name="__module.model.gpt_neox.layers.12.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
 			</rt_info>
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16">
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4534" name="__module.model.gpt_neox.layers.18.attention/aten::neg/Multiply" type="Multiply" version="opset1">
+		<layer id="2071" name="__module.model.gpt_neox.layers.12.mlp.dense_4h_to_h/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3054">
+				<port id="2" precision="FP32" names="1826,input.51">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4535" name="Constant_275866" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4536" name="Constant_275871" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2072" name="Constant_37386" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1127219280" size="10240" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4537" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2073" name="__module.model.gpt_neox.layers.12.input_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3048">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4538" name="__module.model.gpt_neox.layers.18.attention/aten::cat/Concat" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="2074" name="Constant_37387" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1127229520" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2075" name="__module.model.gpt_neox.layers.12.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3056">
+				<port id="2" precision="FP32" names="1729,hidden_states.49">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4539" name="self.model.gpt_neox.layers.18.attention.rotary_emb.sin_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201162356" size="81920" />
+		<layer id="2076" name="self.model.gpt_neox.layers.12.attention.query_key_value.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 2560" offset="1127239760" size="19660800" />
 			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.18.attention.rotary_emb.sin_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4540" name="self.model.gpt_neox.layers.18.attention.rotary_emb.sin_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="2077" name="Convert_74956" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4541" name="__module.model.gpt_neox.layers.18.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2078" name="self.model.gpt_neox.layers.12.attention.query_key_value.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 1" offset="1146900560" size="7680" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="U8">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4542" name="__module.model.gpt_neox.layers.18.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2079" name="Convert_74959" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>1</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4543" name="__module.model.gpt_neox.layers.18.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2080" name="self.model.gpt_neox.layers.12.attention.query_key_value.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2081" name="self.model.gpt_neox.layers.12.attention.query_key_value.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="7680, 1" offset="1146908240" size="15360" />
+			<output>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I64">
+			</output>
+		</layer>
+		<layer id="2082" name="self.model.gpt_neox.layers.12.attention.query_key_value.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3033,3036,sin.73">
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="2" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4544" name="Constant_38308" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
+		<layer id="2083" name="self.model.gpt_neox.layers.12.attention.query_key_value.weight/fq_weights_1/convert" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I32" />
+				<port id="1" precision="FP32">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4545" name="__module.model.gpt_neox.layers.18.attention/aten::index/Gather_1" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="2084" name="__module.model.gpt_neox.layers.12.attention.query_key_value/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="3041">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4546" name="__module.model.gpt_neox.layers.18.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64" />
-			</input>
+		<layer id="2085" name="Constant_37388" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 7680" offset="1146923600" size="30720" />
 			<output>
-				<port id="2" precision="FP32" names="3042,sin.75">
-					<dim>-1</dim>
+				<port id="0" precision="FP32">
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>1</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4547" name="__module.model.gpt_neox.layers.18.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
+		<layer id="2086" name="__module.model.gpt_neox.layers.12.attention.query_key_value/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>7680</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>1</dim>
+					<dim>7680</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3057">
+				<port id="2" precision="FP32" names="1736,qkv.49">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4548" name="__module.model.gpt_neox.layers.18.attention/aten::add/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2087" name="__module.model.gpt_neox.layers.12.attention/prim::ListConstruct" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201115992" size="32" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2088" name="__module.model.gpt_neox.layers.12.attention/aten::view/Reshape" type="Reshape" version="opset1">
+			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>7680</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3058">
+				<port id="2" precision="FP32" names="1740,qkv.51">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4549" name="Constant_275917" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2089" name="Constant_8276" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4550" name="Constant_275920" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2090" name="Constant_8278" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4551" name="Constant_275923" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2091" name="Constant_8280" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4552" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2092" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Slice" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3020">
+				<port id="5" precision="FP32" names="1741">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>60</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4553" name="__module.model.gpt_neox.layers.18.attention/aten::cat/Concat_1" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="2093" name="Constant_8289" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
+			<output>
+				<port id="0" precision="I64" names="1742">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2094" name="__module.model.gpt_neox.layers.12.attention/aten::permute/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3076,3086,query.111,query.113">
+				<port id="2" precision="FP32" names="1743,query.25">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -69586,41 +31565,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4554" name="Constant_275929" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2095" name="Constant_8330" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4555" name="Constant_275932" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2096" name="Constant_8332" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4556" name="Constant_275935" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2097" name="Constant_8334" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4557" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2098" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Slice_1" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -69629,17 +31598,20 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3021,k.37">
+				<port id="5" precision="FP32" names="1750,q.25">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -69647,7 +31619,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4558" name="__module.model.gpt_neox.layers.18.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
+		<layer id="2099" name="__module.model.gpt_neox.layers.12.attention/aten::mul/Multiply" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -69664,7 +31636,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3059">
+				<port id="2" precision="FP32" names="1756">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -69672,160 +31644,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4559" name="Constant_275939" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4560" name="Constant_275938" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2100" name="Constant_8488" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4561" name="__module.model.gpt_neox.layers.18.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4562" name="__module.model.gpt_neox.layers.18.attention/aten::size/Constant_7" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2101" name="Constant_8490" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="4563" name="__module.model.gpt_neox.layers.18.attention/aten::size/Gather_7" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="3060,3065" />
-			</output>
-		</layer>
-		<layer id="4564" name="__module.model.gpt_neox.layers.18.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" />
-			</output>
-		</layer>
-		<layer id="4565" name="__module.model.gpt_neox.layers.18.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-			</input>
-			<output>
-				<port id="1" precision="I64" names="3062,3063,3067,3068" />
-			</output>
-		</layer>
-		<layer id="4566" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="4567" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4568" name="Constant_275937" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2102" name="Constant_8492" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4569" name="ScatterUpdate_275940" type="ScatterUpdate" version="opset3">
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4570" name="Constant_275943" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4571" name="Constant_275946" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
 			</output>
 		</layer>
-		<layer id="4572" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2103" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Slice_2" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -69834,17 +31677,20 @@
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3069,x2.75">
+				<port id="5" precision="FP32" names="1766,x2.49">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -69852,32 +31698,10 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4573" name="Constant_299257_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4574" name="Constant_299257" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
+		<layer id="2104" name="Constant_37389" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>1</dim>
@@ -69885,7 +31709,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4575" name="__module.model.gpt_neox.layers.18.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
+		<layer id="2105" name="__module.model.gpt_neox.layers.12.attention/aten::neg/Multiply" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -69902,7 +31726,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3070">
+				<port id="2" precision="FP32" names="1767">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -69910,30 +31734,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4576" name="Constant_275991" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2106" name="Constant_8385" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4577" name="Constant_275996" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2107" name="Constant_8429" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2108" name="Constant_8431" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4578" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2109" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Slice_3" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -69942,17 +31767,20 @@
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3064">
+				<port id="5" precision="FP32" names="1761">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -69960,7 +31788,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4579" name="__module.model.gpt_neox.layers.18.attention/aten::cat/Concat_2" type="Concat" version="opset1">
+		<layer id="2110" name="__module.model.gpt_neox.layers.12.attention/aten::cat/Concat" type="Concat" version="opset1">
 			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
@@ -69977,7 +31805,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3072">
+				<port id="2" precision="FP32" names="1769">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -69985,7 +31813,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4580" name="__module.model.gpt_neox.layers.18.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
+		<layer id="2111" name="__module.model.gpt_neox.layers.12.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -70002,7 +31830,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3073">
+				<port id="2" precision="FP32" names="1770">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -70010,7 +31838,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4581" name="__module.model.gpt_neox.layers.18.attention/aten::add/Add_1" type="Add" version="opset1">
+		<layer id="2112" name="__module.model.gpt_neox.layers.12.attention/aten::add/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -70027,7 +31855,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3074">
+				<port id="2" precision="FP32" names="1771">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -70035,41 +31863,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4582" name="Constant_276041" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2113" name="Constant_8340" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4583" name="Constant_276044" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2114" name="Constant_8342" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4584" name="Constant_276047" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2115" name="Constant_8344" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4585" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2116" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Slice_4" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -70078,17 +31896,20 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3022">
+				<port id="5" precision="FP32" names="1751">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -70096,7 +31917,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4586" name="__module.model.gpt_neox.layers.18.attention/aten::cat/Concat_3" type="Concat" version="opset1">
+		<layer id="2117" name="__module.model.gpt_neox.layers.12.attention/aten::cat/Concat_1" type="Concat" version="opset1">
 			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
@@ -70113,32 +31934,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3078">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4587" name="__module.model.gpt_neox.layers.18.attention/aten::to/Convert_1" type="Concat" version="opset1">
-			<data axis="-2" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="present.18.key">
+				<port id="2" precision="FP32" names="1789">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -70146,49 +31942,37 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4588" name="Constant_337910_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="4589" name="Constant_337910" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
+		<layer id="2118" name="Constant_52839" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="1" precision="FP32" />
+				<port id="0" precision="FP32" />
 			</output>
 		</layer>
-		<layer id="4590" name="Constant_337903" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
+		<layer id="2119" name="Constant_52832" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4591" name="Constant_337905" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
+		<layer id="2120" name="Constant_52834" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4592" name="Constant_337907" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
+		<layer id="2121" name="Constant_52836" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4593" name="Concat_337908" type="Concat" version="opset1">
+		<layer id="2122" name="Concat_52837" type="Concat" version="opset1">
 			<data axis="0" />
 			<input>
 				<port id="0" precision="I64">
@@ -70210,7 +31994,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4594" name="Broadcast_337911" type="Broadcast" version="opset3">
+		<layer id="2123" name="Broadcast_52840" type="Broadcast" version="opset3">
 			<data mode="numpy" />
 			<input>
 				<port id="0" precision="FP32" />
@@ -70227,8 +32011,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4595" name="ReadValue_330467" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.18.valuepresent.18.value" variable_type="f32" variable_shape="?,32,?,80" />
+		<layer id="2124" name="ReadValue_50199" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.12.keypresent.12.key" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -70238,7 +32022,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="past_key_values.18.value">
+				<port id="1" precision="FP32" names="past_key_values.12.key">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -70246,13 +32030,13 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4596" name="Constant_323496" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
+		<layer id="2125" name="Constant_47850" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="4597" name="Gather_323497" type="Gather" version="opset8">
+		<layer id="2126" name="Gather_47851" type="Gather" version="opset8">
 			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="FP32">
@@ -70275,41 +32059,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4598" name="Constant_276053" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2127" name="Constant_8294" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4599" name="Constant_276056" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2128" name="Constant_8296" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4600" name="Constant_276059" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2129" name="Constant_8298" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4601" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2130" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Slice_5" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -70318,17 +32092,20 @@
 					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3016">
+				<port id="5" precision="FP32" names="1744">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -70336,15 +32113,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4602" name="Constant_10709" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="2131" name="Constant_8307" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64" names="3017">
+				<port id="0" precision="I64" names="1745">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4603" name="__module.model.gpt_neox.layers.18.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
+		<layer id="2132" name="__module.model.gpt_neox.layers.12.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -70357,7 +32134,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3018">
+				<port id="2" precision="FP32" names="1746,key.49">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -70365,8 +32142,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4604" name="__module.model.gpt_neox.layers.18.attention/aten::cat/Concat_5" type="Concat" version="opset1">
-			<data axis="-2" />
+		<layer id="2133" name="Constant_8350" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2134" name="Constant_8352" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2135" name="Constant_8354" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2136" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Slice_6" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -70374,176 +32174,303 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="5" precision="FP32" names="1752,k.25">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2137" name="__module.model.gpt_neox.layers.12.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="present.18.value">
+				<port id="2" precision="FP32" names="1772">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4605" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Slice_11" type="Const" version="opset1">
-			<data element_type="boolean" shape="1, 1, 2048, 2048" offset="201244276" size="4194304" />
+		<layer id="2138" name="Constant_8631" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
-				<port id="0" precision="BOOL" names="3095">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4606" name="Constant_276063" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2139" name="Constant_8633" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>3</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4607" name="Constant_276062" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162298" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2140" name="Constant_8635" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4608" name="__module.model.gpt_neox.layers.18.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="2141" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Slice_7" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="5" precision="FP32" names="1782,x2.51">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4609" name="__module.model.gpt_neox.layers.18.attention/aten::size/Constant_8" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2142" name="Constant_37390" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4610" name="__module.model.gpt_neox.layers.18.attention/aten::size/Gather_8" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="2143" name="__module.model.gpt_neox.layers.12.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="3090" />
+				<port id="2" precision="FP32" names="1783">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4611" name="__module.model.gpt_neox.layers.18.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="2144" name="Constant_8528" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2145" name="Constant_8572" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2146" name="Constant_8574" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2147" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Slice_8" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="5" precision="FP32" names="1777">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4612" name="__module.model.gpt_neox.layers.18.attention/aten::size/Constant_9" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2148" name="__module.model.gpt_neox.layers.12.attention/aten::cat/Concat_2" type="Concat" version="opset1">
+			<data axis="-1" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I32" />
+				<port id="2" precision="FP32" names="1785">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4613" name="__module.model.gpt_neox.layers.18.attention/aten::size/Gather_9" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="2149" name="__module.model.gpt_neox.layers.12.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="3088" />
+				<port id="2" precision="FP32" names="1786">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4614" name="__module.model.gpt_neox.layers.18.attention/aten::sub/Subtract" type="Subtract" version="opset1">
+		<layer id="2150" name="__module.model.gpt_neox.layers.12.attention/aten::add/Add_1" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64" names="3092,3093" />
+				<port id="2" precision="FP32" names="1787">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4615" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Constant_4" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2151" name="Constant_8360" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4616" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="2152" name="Constant_8362" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4617" name="Constant_276061" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2153" name="Constant_8364" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4618" name="ScatterUpdate_276064" type="ScatterUpdate" version="opset3">
+		<layer id="2154" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Slice_9" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -70551,280 +32478,313 @@
 				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
 				<port id="4" precision="I64">
-					<dim>3</dim>
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="4619" name="Constant_276065" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="5" precision="FP32" names="1753">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4620" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1">
+		<layer id="2155" name="__module.model.gpt_neox.layers.12.attention/aten::cat/Concat_3" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="1791">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4621" name="ScatterUpdate_276066" type="ScatterUpdate" version="opset3">
+		<layer id="2156" name="__module.model.gpt_neox.layers.12.attention/aten::cat/Concat_4" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="present.12.key">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="3" precision="I32">
+			</output>
+		</layer>
+		<layer id="2157" name="Constant_52851" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
+			<output>
+				<port id="0" precision="FP32" />
+			</output>
+		</layer>
+		<layer id="2158" name="Constant_52844" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="2159" name="Constant_52846" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4622" name="Constant_276069" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438612" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2160" name="Constant_52848" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>3</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4623" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2161" name="Concat_52849" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>3</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>3</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>3</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="BOOL" names="3096">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
+				<port id="4" precision="I64">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4624" name="Constant_276142" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
+		<layer id="2162" name="Broadcast_52852" type="Broadcast" version="opset3">
+			<data mode="numpy" />
+			<input>
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="4625" name="Constant_276143" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4626" name="Constant_276139" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2163" name="ReadValue_50201" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.12.valuepresent.12.value" variable_type="f32" variable_shape="?,32,?,80" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="1" precision="FP32" names="past_key_values.12.value">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4627" name="Constant_276138" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2164" name="Constant_47853" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="4628" name="ScatterUpdate_276144" type="ScatterUpdate" version="opset3">
+		<layer id="2165" name="Gather_47854" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
+				<port id="1" precision="I32">
+					<dim>-1</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="2" precision="I64" />
+			</input>
+			<output>
+				<port id="3" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="3" precision="I32">
+			</output>
+		</layer>
+		<layer id="2166" name="Constant_8312" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="2167" name="Constant_8314" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4629" name="Constant_276147" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2168" name="Constant_8316" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4630" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2169" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Slice_10" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>2048</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="BOOL" names="3097,3098">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="5" precision="FP32" names="1747">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4631" name="__module.model.gpt_neox.layers.18.attention/aten::where/Convert_1_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="4632" name="__module.model.gpt_neox.layers.18.attention/aten::where/Convert_1" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="4633" name="__module.model.gpt_neox.layers.18.attention/aten::full/Convert" type="Const" version="opset1">
-			<data element_type="f32" shape="" offset="205438636" size="4" />
+		<layer id="2170" name="Constant_8325" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="FP32" names="3085" />
+				<port id="0" precision="I64" names="1748">
+					<dim>4</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4634" name="__module.model.gpt_neox.layers.18.attention/aten::where/Select" type="Select" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2171" name="__module.model.gpt_neox.layers.12.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
 			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
-				<port id="1" precision="FP32" />
-				<port id="2" precision="FP32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="3099,causal_mask.73">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="1749">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4635" name="Add_290446" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2172" name="__module.model.gpt_neox.layers.12.attention/aten::cat/Concat_5" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="present.12.value">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2173" name="Constant_8707" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4636" name="ShapeOf_290447" type="ShapeOf" version="opset3">
+		<layer id="2174" name="ShapeOf_39564" type="ShapeOf" version="opset3">
 			<data output_type="i64" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
@@ -70833,27 +32793,21 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4637" name="Constant_254599" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2175" name="Constant_39565" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116168" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4638" name="__module.model.gpt_neox.layers.18.attention/aten::size/Constant_10" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2176" name="Constant_39566" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="4639" name="__module.model.gpt_neox.layers.18.attention/aten::size/Gather_10" type="Gather" version="opset8">
+		<layer id="2177" name="Gather_39567" type="Gather" version="opset8">
 			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="I64">
@@ -70862,16 +32816,16 @@
 				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I32" />
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="3083">
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4640" name="__module.model.gpt_neox.layers.18.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="2178" name="Add_39570" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="I64">
 					<dim>1</dim>
@@ -70879,49 +32833,22 @@
 				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64" names="3100">
-					<dim>4</dim>
+				<port id="2" precision="I64" names="1796">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4641" name="__module.model.gpt_neox.layers.18.attention/aten::expand/Abs" type="Abs" version="opset1">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
+		<layer id="2179" name="Constant_8710" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4642" name="Maximum_290448" type="Maximum" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4643" name="__module.model.gpt_neox.layers.18.attention/aten::add/Add_2" type="Broadcast" version="opset3">
-			<data mode="numpy" />
+		<layer id="2180" name="__module.model.gpt_neox.layers.12.attention/aten::slice/Slice_14" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -70930,19 +32857,28 @@
 					<dim>-1</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3102">
-					<dim>-1</dim>
+				<port id="5" precision="FP32" names="1800">
 					<dim>-1</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
 					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4644" name="__module.model.gpt_neox.layers.18.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+		<layer id="2181" name="__module.model.gpt_neox.layers.12.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
 			<data causal="false" />
 			<input>
 				<port id="0" precision="FP32">
@@ -70965,13 +32901,13 @@
 				</port>
 				<port id="3" precision="FP32">
 					<dim>-1</dim>
-					<dim>-1</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
 					<dim>-1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3103,3104,sdpa_result.37,tensor.73">
+				<port id="4" precision="FP32" names="1801,attn_output.73">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -70979,15 +32915,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4645" name="Constant_11150" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="2182" name="__module.model.gpt_neox.layers.12.attention/aten::transpose/Constant" type="Const" version="opset1">
+			<data element_type="i32" shape="4" offset="201116200" size="16" />
 			<output>
-				<port id="0" precision="I64" names="3105">
+				<port id="0" precision="I32">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4646" name="__module.model.gpt_neox.layers.18.attention/aten::permute/Transpose_3" type="Transpose" version="opset1">
+		<layer id="2183" name="__module.model.gpt_neox.layers.12.attention/aten::transpose/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -70995,12 +32931,12 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="I32">
 					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3106">
+				<port id="2" precision="FP32" names="1802">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -71008,18 +32944,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4647" name="__module.model.gpt_neox.layers.18.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438650" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2184" name="Constant_40996" type="Const" version="opset1">
+			<data element_type="i64" shape="3" offset="201116216" size="24" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>3</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4648" name="__module.model.gpt_neox.layers.18.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
+		<layer id="2185" name="__module.model.gpt_neox.layers.12.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
 			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -71033,15 +32966,15 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3111">
+				<port id="2" precision="FP32" names="1805">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4649" name="self.model.gpt_neox.layers.18.attention.dense.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 2560" offset="1623458514" size="6553600" />
+		<layer id="2186" name="self.model.gpt_neox.layers.12.attention.dense.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 2560" offset="1146954320" size="6553600" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>2560</dim>
@@ -71049,7 +32982,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4650" name="Convert_22154" type="Convert" version="opset1">
+		<layer id="2187" name="Convert_75308" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -71064,8 +32997,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4651" name="self.model.gpt_neox.layers.18.attention.dense.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="1630012114" size="2560" />
+		<layer id="2188" name="self.model.gpt_neox.layers.12.attention.dense.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="1153507920" size="2560" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>2560</dim>
@@ -71073,7 +33006,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4652" name="Convert_22157" type="Convert" version="opset1">
+		<layer id="2189" name="Convert_75311" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -71088,7 +33021,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4653" name="self.model.gpt_neox.layers.18.attention.dense.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="2190" name="self.model.gpt_neox.layers.12.attention.dense.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -71107,8 +33040,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4654" name="self.model.gpt_neox.layers.18.attention.dense.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="1630014674" size="5120" />
+		<layer id="2191" name="self.model.gpt_neox.layers.12.attention.dense.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="1153510480" size="5120" />
 			<output>
 				<port id="0" precision="FP16">
 					<dim>2560</dim>
@@ -71116,7 +33049,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4655" name="self.model.gpt_neox.layers.18.attention.dense.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="2192" name="self.model.gpt_neox.layers.12.attention.dense.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -71135,11 +33068,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4656" name="self.model.gpt_neox.layers.18.attention.dense.weight" type="Convert" version="opset1">
+		<layer id="2193" name="self.model.gpt_neox.layers.12.attention.dense.weight/fq_weights_1/convert" type="Convert" version="opset1">
 			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
 			<input>
 				<port id="0" precision="FP16">
 					<dim>2560</dim>
@@ -71153,7 +33083,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4657" name="__module.model.gpt_neox.layers.18.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="2194" name="__module.model.gpt_neox.layers.12.attention.dense/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -71174,37 +33104,121 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4658" name="Constant_299258_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1630019794" size="5120" />
+		<layer id="2195" name="Constant_37391" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1153515600" size="10240" />
 			<output>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4659" name="Constant_299258" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="2196" name="__module.model.gpt_neox.layers.12.attention.dense/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
+				<port id="2" precision="FP32" names="1808,1810,input.49">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2197" name="__module.model.gpt_neox.layers.12/aten::add/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="1828">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2198" name="__module.model.gpt_neox.layers.12/aten::add/Add_1" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
 				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="1829,1831,hidden_states.55">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2199" name="__module.model.gpt_neox.layers.13.input_layernorm/aten::layer_norm/Multiply" type="Const" version="opset1">
+			<data element_type="i32" shape="1" offset="128867668" size="4" />
+			<output>
+				<port id="0" precision="I32">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2200" name="__module.model.gpt_neox.layers.13.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
+			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="I32">
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2201" name="Constant_37392" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1153525840" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4660" name="__module.model.gpt_neox.layers.18.attention.dense/aten::linear/Add" type="Add" version="opset1">
+		<layer id="2202" name="__module.model.gpt_neox.layers.13.post_attention_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -71219,14 +33233,24 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3114,3116,input.75">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4661" name="__module.model.gpt_neox.layers.18/aten::add/Add" type="Add" version="opset1">
+		<layer id="2203" name="Constant_37393" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1153536080" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2204" name="__module.model.gpt_neox.layers.13.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -71235,59 +33259,338 @@
 					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="1929">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
+			</output>
+		</layer>
+		<layer id="2205" name="self.model.gpt_neox.layers.13.mlp.dense_h_to_4h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 2560" offset="1153546320" size="26214400" />
+			<output>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2206" name="Convert_74444" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2207" name="self.model.gpt_neox.layers.13.mlp.dense_h_to_4h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 1" offset="1179760720" size="10240" />
+			<output>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2208" name="Convert_74447" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2209" name="self.model.gpt_neox.layers.13.mlp.dense_h_to_4h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2210" name="self.model.gpt_neox.layers.13.mlp.dense_h_to_4h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="10240, 1" offset="1179770960" size="20480" />
+			<output>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2211" name="self.model.gpt_neox.layers.13.mlp.dense_h_to_4h.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2212" name="__module.model.gpt_neox.layers.13.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<rt_info>
+				<attribute name="decompression" version="0" />
+			</rt_info>
+			<input>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3134">
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2213" name="__module.model.gpt_neox.layers.13.mlp.dense_h_to_4h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
+			<input>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2214" name="self.model.gpt_neox.layers.13.mlp.dense_h_to_4h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="10240" offset="1179791440" size="20480" />
+			<output>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.13.mlp.dense_h_to_4h.bias">
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2215" name="__module.model.gpt_neox.layers.13.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<rt_info>
+				<attribute name="decompression" version="0" />
+			</rt_info>
+			<input>
+				<port id="0" precision="BF16">
+					<dim>10240</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4662" name="__module.model.gpt_neox.layers.18/aten::add/Add_1" type="Add" version="opset1">
+		<layer id="2216" name="__module.model.gpt_neox.layers.13.mlp.dense_h_to_4h/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 				<port id="1" precision="FP32">
+					<dim>10240</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="1934">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2217" name="__module.model.gpt_neox.layers.13.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
+			<data approximation_mode="ERF" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3135,3137,hidden_states.39">
+				<port id="1" precision="FP32" names="1935">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2218" name="self.model.gpt_neox.layers.13.mlp.dense_4h_to_h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 10240" offset="1179811920" size="26214400" />
+			<output>
+				<port id="0" precision="U8">
 					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4663" name="__module.model.gpt_neox.layers.19.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867670" size="4" />
+		<layer id="2219" name="Convert_74454" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I32">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2220" name="self.model.gpt_neox.layers.13.mlp.dense_4h_to_h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="1206026320" size="2560" />
+			<output>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4664" name="__module.model.gpt_neox.layers.19.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
-			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
+		<layer id="2221" name="Convert_74457" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2222" name="self.model.gpt_neox.layers.13.mlp.dense_4h_to_h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2223" name="self.model.gpt_neox.layers.13.mlp.dense_4h_to_h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="1206028880" size="5120" />
+			<output>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2224" name="self.model.gpt_neox.layers.13.mlp.dense_4h_to_h.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2225" name="__module.model.gpt_neox.layers.13.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<rt_info>
+				<attribute name="decompression" version="0" />
+			</rt_info>
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2226" name="__module.model.gpt_neox.layers.13.mlp.dense_4h_to_h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="1" precision="I32">
-					<dim>1</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
@@ -71298,37 +33601,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4665" name="Constant_299259_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1630024914" size="5120" />
+		<layer id="2227" name="self.model.gpt_neox.layers.13.mlp.dense_4h_to_h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="2560" offset="1206034000" size="5120" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.13.mlp.dense_4h_to_h.bias">
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4666" name="Constant_299259" type="Convert" version="opset1">
+		<layer id="2228" name="__module.model.gpt_neox.layers.13.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
 			</rt_info>
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16">
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4667" name="__module.model.gpt_neox.layers.19.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
+		<layer id="2229" name="__module.model.gpt_neox.layers.13.mlp.dense_4h_to_h/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -71337,50 +33634,60 @@
 					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="1938,input.55">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4668" name="Constant_299260_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1630030034" size="5120" />
+		<layer id="2230" name="Constant_37394" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1206039120" size="10240" />
 			<output>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4669" name="Constant_299260" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="2231" name="__module.model.gpt_neox.layers.13.input_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2232" name="Constant_37395" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1206049360" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4670" name="__module.model.gpt_neox.layers.19.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+		<layer id="2233" name="__module.model.gpt_neox.layers.13.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -71395,127 +33702,124 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3272">
+				<port id="2" precision="FP32" names="1841,hidden_states.53">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4671" name="self.model.gpt_neox.layers.19.mlp.dense_h_to_4h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 2560" offset="1630035154" size="26214400" />
+		<layer id="2234" name="self.model.gpt_neox.layers.13.attention.query_key_value.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 2560" offset="1206059600" size="19660800" />
 			<output>
 				<port id="0" precision="U8">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4672" name="Convert_22204" type="Convert" version="opset1">
+		<layer id="2235" name="Convert_74967" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP16">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4673" name="self.model.gpt_neox.layers.19.mlp.dense_h_to_4h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 1" offset="1656249554" size="10240" />
+		<layer id="2236" name="self.model.gpt_neox.layers.13.attention.query_key_value.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 1" offset="1225720400" size="7680" />
 			<output>
 				<port id="0" precision="U8">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4674" name="Convert_22207" type="Convert" version="opset1">
+		<layer id="2237" name="Convert_74970" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP16">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4675" name="self.model.gpt_neox.layers.19.mlp.dense_h_to_4h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="2238" name="self.model.gpt_neox.layers.13.attention.query_key_value.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP16">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
 				<port id="2" precision="FP16">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4676" name="self.model.gpt_neox.layers.19.mlp.dense_h_to_4h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="10240, 1" offset="1656259794" size="20480" />
+		<layer id="2239" name="self.model.gpt_neox.layers.13.attention.query_key_value.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="7680, 1" offset="1225728080" size="15360" />
 			<output>
 				<port id="0" precision="FP16">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4677" name="self.model.gpt_neox.layers.19.mlp.dense_h_to_4h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="2240" name="self.model.gpt_neox.layers.13.attention.query_key_value.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP16">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
 				<port id="2" precision="FP16">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4678" name="self.model.gpt_neox.layers.19.mlp.dense_h_to_4h.weight" type="Convert" version="opset1">
+		<layer id="2241" name="self.model.gpt_neox.layers.13.attention.query_key_value.weight/fq_weights_1/convert" type="Convert" version="opset1">
 			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
 			<input>
 				<port id="0" precision="FP16">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP32">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4679" name="__module.model.gpt_neox.layers.19.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="2242" name="__module.model.gpt_neox.layers.13.attention.query_key_value/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -71524,7 +33828,7 @@
 					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
@@ -71532,859 +33836,1101 @@
 				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4680" name="Constant_299261_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 10240" offset="1656280274" size="20480" />
+		<layer id="2243" name="Constant_37396" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 7680" offset="1225743440" size="30720" />
 			<output>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>10240</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4681" name="Constant_299261" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="2244" name="__module.model.gpt_neox.layers.13.attention.query_key_value/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>7680</dim>
+				</port>
+				<port id="1" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>10240</dim>
+					<dim>7680</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
+				<port id="2" precision="FP32" names="1848,qkv.53">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>7680</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2245" name="__module.model.gpt_neox.layers.13.attention/prim::ListConstruct" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201115992" size="32" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2246" name="__module.model.gpt_neox.layers.13.attention/aten::view/Reshape" type="Reshape" version="opset1">
+			<data special_zero="true" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>7680</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>4</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="1852,qkv.55">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2247" name="Constant_8906" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2248" name="Constant_8908" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4682" name="__module.model.gpt_neox.layers.19.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2249" name="Constant_8910" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2250" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Slice" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
-				<port id="1" precision="FP32">
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3277">
+				<port id="5" precision="FP32" names="1853">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4683" name="__module.model.gpt_neox.layers.19.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
-			<data approximation_mode="ERF" />
+		<layer id="2251" name="Constant_8919" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
+			<output>
+				<port id="0" precision="I64" names="1854">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2252" name="__module.model.gpt_neox.layers.13.attention/aten::permute/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="3278">
+				<port id="2" precision="FP32" names="1855,query.27">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4684" name="self.model.gpt_neox.layers.19.mlp.dense_4h_to_h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 10240" offset="1656300754" size="26214400" />
+		<layer id="2253" name="Constant_8960" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4685" name="Convert_22214" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
+		<layer id="2254" name="Constant_8962" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4686" name="self.model.gpt_neox.layers.19.mlp.dense_4h_to_h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="1682515154" size="2560" />
+		<layer id="2255" name="Constant_8964" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4687" name="Convert_22217" type="Convert" version="opset1">
-			<data destination_type="f16" />
+		<layer id="2256" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Slice_1" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
+				<port id="5" precision="FP32" names="1862,q.27">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4688" name="self.model.gpt_neox.layers.19.mlp.dense_4h_to_h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="2257" name="__module.model.gpt_neox.layers.13.attention/aten::mul/Multiply" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
 					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="2" precision="FP32" names="1868">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4689" name="self.model.gpt_neox.layers.19.mlp.dense_4h_to_h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="1682517714" size="5120" />
+		<layer id="2258" name="Constant_9118" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4690" name="self.model.gpt_neox.layers.19.mlp.dense_4h_to_h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+		<layer id="2259" name="Constant_9120" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="2260" name="Constant_9122" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4691" name="self.model.gpt_neox.layers.19.mlp.dense_4h_to_h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="2261" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Slice_2" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="5" precision="FP32" names="1878,x2.53">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4692" name="__module.model.gpt_neox.layers.19.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
+		<layer id="2262" name="Constant_37397" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2263" name="__module.model.gpt_neox.layers.13.attention/aten::neg/Multiply" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>10</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="1879">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4693" name="Constant_299262_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1682522834" size="5120" />
+		<layer id="2264" name="Constant_9015" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="FP16">
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2265" name="Constant_9059" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4694" name="Constant_299262" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="2266" name="Constant_9061" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2267" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Slice_3" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
+				<port id="5" precision="FP32" names="1873">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2268" name="__module.model.gpt_neox.layers.13.attention/aten::cat/Concat" type="Concat" version="opset1">
+			<data axis="-1" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="1881">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4695" name="__module.model.gpt_neox.layers.19.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1">
+		<layer id="2269" name="__module.model.gpt_neox.layers.13.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
+					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3281,input.81">
+				<port id="2" precision="FP32" names="1882">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4696" name="Constant_299263_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1682527954" size="5120" />
+		<layer id="2270" name="__module.model.gpt_neox.layers.13.attention/aten::add/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="FP16">
+				<port id="2" precision="FP32" names="1883">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2271" name="Constant_8970" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2272" name="Constant_8972" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4697" name="Constant_299263" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="2273" name="Constant_8974" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2274" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Slice_4" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+				<port id="5" precision="FP32" names="1863">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4698" name="__module.model.gpt_neox.layers.19.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2275" name="__module.model.gpt_neox.layers.13.attention/aten::cat/Concat_1" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="1901">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2276" name="Constant_52863" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
+			<output>
+				<port id="0" precision="FP32" />
+			</output>
+		</layer>
+		<layer id="2277" name="Constant_52856" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2278" name="Constant_52858" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2279" name="Constant_52860" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2280" name="Concat_52861" type="Concat" version="opset1">
+			<data axis="0" />
+			<input>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4699" name="Constant_299264_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1682533074" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+				<port id="4" precision="I64">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4700" name="Constant_299264" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="2281" name="Broadcast_52864" type="Broadcast" version="opset3">
+			<data mode="numpy" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4701" name="__module.model.gpt_neox.layers.19.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2282" name="ReadValue_50203" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.13.keypresent.13.key" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3147">
+				<port id="1" precision="FP32" names="past_key_values.13.key">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4702" name="self.model.gpt_neox.layers.19.attention.query_key_value.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 2560" offset="1682538194" size="19660800" />
+		<layer id="2283" name="Constant_47856" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="4703" name="Convert_22184" type="Convert" version="opset1">
-			<data destination_type="f16" />
+		<layer id="2284" name="Gather_47857" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I32">
+					<dim>-1</dim>
 				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="3" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4704" name="self.model.gpt_neox.layers.19.attention.query_key_value.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 1" offset="1702198994" size="7680" />
+		<layer id="2285" name="Constant_8924" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4705" name="Convert_22187" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
+		<layer id="2286" name="Constant_8926" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="2287" name="Constant_8928" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4706" name="self.model.gpt_neox.layers.19.attention.query_key_value.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2288" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Slice_5" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
+				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4707" name="self.model.gpt_neox.layers.19.attention.query_key_value.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="7680, 1" offset="1702206674" size="15360" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
+				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="4708" name="self.model.gpt_neox.layers.19.attention.query_key_value.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="3" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="5" precision="FP32" names="1856">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4709" name="self.model.gpt_neox.layers.19.attention.query_key_value.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="2289" name="Constant_8937" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64" names="1857">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4710" name="__module.model.gpt_neox.layers.19.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
+		<layer id="2290" name="__module.model.gpt_neox.layers.13.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="1858,key.53">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4711" name="Constant_299265_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 7680" offset="1702222034" size="15360" />
+		<layer id="2291" name="Constant_8980" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4712" name="Constant_299265" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+		<layer id="2292" name="Constant_8982" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>7680</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="2293" name="Constant_8984" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4713" name="__module.model.gpt_neox.layers.19.attention.query_key_value/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2294" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Slice_6" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>7680</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3154,qkv.77">
+				<port id="5" precision="FP32" names="1864,k.27">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4714" name="__module.model.gpt_neox.layers.19.attention/prim::ListConstruct/Concat" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080154" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4715" name="__module.model.gpt_neox.layers.19.attention/aten::view/Reshape" type="Reshape" version="opset1">
-			<data special_zero="true" />
+		<layer id="2295" name="__module.model.gpt_neox.layers.13.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3158,qkv.79">
-					<dim>-1</dim>
+				<port id="2" precision="FP32" names="1884">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>240</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4716" name="Constant_276154" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2296" name="Constant_9261" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4717" name="Constant_276157" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2297" name="Constant_9263" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4718" name="Constant_276160" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2298" name="Constant_9265" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4719" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2299" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Slice_7" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>240</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3159">
-					<dim>-1</dim>
+				<port id="5" precision="FP32" names="1894,x2.55">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4720" name="Constant_11262" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="2300" name="Constant_37398" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="0" precision="I64" names="3160">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4721" name="__module.model.gpt_neox.layers.19.attention/aten::permute/Transpose" type="Transpose" version="opset1">
+		<layer id="2301" name="__module.model.gpt_neox.layers.13.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3161,query.115">
+				<port id="2" precision="FP32" names="1895">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4722" name="Constant_276166" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2302" name="Constant_9158" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4723" name="Constant_276169" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2303" name="Constant_9202" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4724" name="Constant_276172" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2304" name="Constant_9204" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4725" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2305" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Slice_8" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3168,q.39">
+				<port id="5" precision="FP32" names="1889">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4726" name="self.model.gpt_neox.layers.19.attention.rotary_emb.cos_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201080346" size="81920" />
+		<layer id="2306" name="__module.model.gpt_neox.layers.13.attention/aten::cat/Concat_2" type="Concat" version="opset1">
+			<data axis="-1" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.19.attention.rotary_emb.cos_cached">
-					<dim>2048</dim>
+				<port id="2" precision="FP32" names="1897">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4727" name="self.model.gpt_neox.layers.19.attention.rotary_emb.cos_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="2307" name="__module.model.gpt_neox.layers.13.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
+				<port id="2" precision="FP32" names="1898">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4728" name="__module.model.gpt_neox.layers.19.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2308" name="__module.model.gpt_neox.layers.13.attention/aten::add/Add_1" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="1899">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4729" name="Constant_276178" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2309" name="Constant_8990" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4730" name="Constant_276181" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2310" name="Constant_8992" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4731" name="Constant_276184" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2311" name="Constant_8994" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4732" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2312" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Slice_9" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>240</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3162">
-					<dim>-1</dim>
+				<port id="5" precision="FP32" names="1865">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4733" name="Constant_11276" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="3163">
-					<dim>4</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4734" name="__module.model.gpt_neox.layers.19.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
+		<layer id="2313" name="__module.model.gpt_neox.layers.13.attention/aten::cat/Concat_3" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3164,key.115">
+				<port id="2" precision="FP32" names="1903">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -72392,8 +34938,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4735" name="__module.model.gpt_neox.layers.19.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="2314" name="__module.model.gpt_neox.layers.13.attention/aten::cat/Concat_4" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -72401,78 +34947,53 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4736" name="__module.model.gpt_neox.layers.19.attention/aten::size/Constant_2" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="4737" name="__module.model.gpt_neox.layers.19.attention/aten::size/Gather_2" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="3172" />
-			</output>
-		</layer>
-		<layer id="4738" name="Constant_337923_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
+				<port id="2" precision="FP32" names="present.13.key">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4739" name="Constant_337923" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
+		<layer id="2315" name="Constant_52875" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="1" precision="FP32" />
+				<port id="0" precision="FP32" />
 			</output>
 		</layer>
-		<layer id="4740" name="Constant_337916" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
+		<layer id="2316" name="Constant_52868" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4741" name="Constant_337918" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
+		<layer id="2317" name="Constant_52870" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4742" name="Constant_337920" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
+		<layer id="2318" name="Constant_52872" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4743" name="Concat_337921" type="Concat" version="opset1">
+		<layer id="2319" name="Concat_52873" type="Concat" version="opset1">
 			<data axis="0" />
 			<input>
 				<port id="0" precision="I64">
@@ -72494,7 +35015,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4744" name="Broadcast_337924" type="Broadcast" version="opset3">
+		<layer id="2320" name="Broadcast_52876" type="Broadcast" version="opset3">
 			<data mode="numpy" />
 			<input>
 				<port id="0" precision="FP32" />
@@ -72511,8 +35032,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4745" name="ReadValue_330469" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.19.keypresent.19.key" variable_type="f32" variable_shape="?,32,?,80" />
+		<layer id="2321" name="ReadValue_50205" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.13.valuepresent.13.value" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -72522,7 +35043,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="past_key_values.19.key">
+				<port id="1" precision="FP32" names="past_key_values.13.value">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -72530,120 +35051,66 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4746" name="Constant_323499" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
+		<layer id="2322" name="Constant_47859" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="4747" name="Gather_323500" type="Gather" version="opset8">
+		<layer id="2323" name="Gather_47860" type="Gather" version="opset8">
 			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I64" />
-			</input>
-			<output>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4748" name="__module.model.gpt_neox.layers.19.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4749" name="__module.model.gpt_neox.layers.19.attention/aten::size/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="4750" name="__module.model.gpt_neox.layers.19.attention/aten::size/Gather_3" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="3174" />
-			</output>
-		</layer>
-		<layer id="4751" name="__module.model.gpt_neox.layers.19.attention/aten::add_/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I32">
+					<dim>-1</dim>
+				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="I64" names="3173,3177,3178,seq_len.77" />
+				<port id="3" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4752" name="__module.model.gpt_neox.layers.19.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2324" name="Constant_8942" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4753" name="__module.model.gpt_neox.layers.19.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="2325" name="Constant_8944" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4754" name="__module.model.gpt_neox.layers.19.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2326" name="Constant_8946" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4755" name="__module.model.gpt_neox.layers.19.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2327" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Slice_10" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -72654,114 +35121,89 @@
 				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3181,3184,cos.77">
+				<port id="5" precision="FP32" names="1859">
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4756" name="Constant_38311" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="4757" name="__module.model.gpt_neox.layers.19.attention/aten::index/Gather" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
+		<layer id="2328" name="Constant_8955" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="3" precision="FP32" names="3187">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="I64" names="1860">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4758" name="__module.model.gpt_neox.layers.19.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1">
+		<layer id="2329" name="__module.model.gpt_neox.layers.13.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
-				<port id="1" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3188,cos.79">
+				<port id="2" precision="FP32" names="1861">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4759" name="__module.model.gpt_neox.layers.19.attention/aten::mul/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2330" name="__module.model.gpt_neox.layers.13.attention/aten::cat/Concat_5" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3192">
+				<port id="2" precision="FP32" names="present.13.value">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4760" name="Constant_276189" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4761" name="Constant_276188" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2331" name="Constant_9337" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4762" name="__module.model.gpt_neox.layers.19.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3">
+		<layer id="2332" name="ShapeOf_39635" type="ShapeOf" version="opset3">
 			<data output_type="i64" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
@@ -72770,1259 +35212,1139 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4763" name="__module.model.gpt_neox.layers.19.attention/aten::size/Constant_5" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2333" name="Constant_39636" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116168" size="8" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="4764" name="__module.model.gpt_neox.layers.19.attention/aten::size/Gather_5" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="3193,3198" />
 			</output>
 		</layer>
-		<layer id="4765" name="__module.model.gpt_neox.layers.19.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
+		<layer id="2334" name="Constant_39637" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="2" precision="I64" />
-			</output>
-		</layer>
-		<layer id="4766" name="__module.model.gpt_neox.layers.19.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1">
-			<input>
 				<port id="0" precision="I64" />
-			</input>
-			<output>
-				<port id="1" precision="I64" names="3195,3196,3200,3201" />
-			</output>
-		</layer>
-		<layer id="4767" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Constant_1" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
 			</output>
 		</layer>
-		<layer id="4768" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1">
+		<layer id="2335" name="Gather_39638" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="I64">
+					<dim>4</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="4769" name="Constant_276187" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+				<port id="2" precision="I64" />
+			</input>
 			<output>
-				<port id="0" precision="I32">
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4770" name="ScatterUpdate_276190" type="ScatterUpdate" version="opset3">
+		<layer id="2336" name="Add_39641" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4771" name="Constant_276193" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="I64" names="1908">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4772" name="Constant_276196" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2337" name="Constant_9340" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4773" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2338" name="__module.model.gpt_neox.layers.13.attention/aten::slice/Slice_14" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="3202,x2.77">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4774" name="Constant_299266_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="4775" name="Constant_299266" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="5" precision="FP32" names="1912">
+					<dim>-1</dim>
 					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4776" name="__module.model.gpt_neox.layers.19.attention/aten::neg/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2339" name="__module.model.gpt_neox.layers.13.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+			<data causal="false" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="3" precision="FP32">
+					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3203">
+				<port id="4" precision="FP32" names="1913,attn_output.79">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4777" name="Constant_276241" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4778" name="Constant_276246" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2340" name="__module.model.gpt_neox.layers.13.attention/aten::transpose/Constant" type="Const" version="opset1">
+			<data element_type="i32" shape="4" offset="201116200" size="16" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="I32">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4779" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2341" name="__module.model.gpt_neox.layers.13.attention/aten::transpose/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>80</dim>
 				</port>
-				<port id="3" precision="I64">
+				<port id="1" precision="I32">
 					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3197">
+				<port id="2" precision="FP32" names="1914">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4780" name="__module.model.gpt_neox.layers.19.attention/aten::cat/Concat" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="2342" name="Constant_40997" type="Const" version="opset1">
+			<data element_type="i64" shape="3" offset="201116216" size="24" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>3</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2343" name="__module.model.gpt_neox.layers.13.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
+			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>3</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3205">
+				<port id="2" precision="FP32" names="1917">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4781" name="self.model.gpt_neox.layers.19.attention.rotary_emb.sin_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201162356" size="81920" />
+		<layer id="2344" name="self.model.gpt_neox.layers.13.attention.dense.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 2560" offset="1225774160" size="6553600" />
 			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.19.attention.rotary_emb.sin_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4782" name="self.model.gpt_neox.layers.19.attention.rotary_emb.sin_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="2345" name="Convert_75319" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4783" name="__module.model.gpt_neox.layers.19.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2346" name="self.model.gpt_neox.layers.13.attention.dense.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="1232327760" size="2560" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="U8">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4784" name="__module.model.gpt_neox.layers.19.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2347" name="Convert_75322" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4785" name="__module.model.gpt_neox.layers.19.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2348" name="self.model.gpt_neox.layers.13.attention.dense.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3182,3185,sin.77">
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4786" name="Constant_38314" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
+		<layer id="2349" name="self.model.gpt_neox.layers.13.attention.dense.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="1232330320" size="5120" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4787" name="__module.model.gpt_neox.layers.19.attention/aten::index/Gather_1" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="2350" name="self.model.gpt_neox.layers.13.attention.dense.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
 				</port>
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="3190">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4788" name="__module.model.gpt_neox.layers.19.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1">
+		<layer id="2351" name="self.model.gpt_neox.layers.13.attention.dense.weight/fq_weights_1/convert" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3191,sin.79">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4789" name="__module.model.gpt_neox.layers.19.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2352" name="__module.model.gpt_neox.layers.13.attention.dense/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3206">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2353" name="Constant_37399" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1232335440" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4790" name="__module.model.gpt_neox.layers.19.attention/aten::add/Add" type="Add" version="opset1">
+		<layer id="2354" name="__module.model.gpt_neox.layers.13.attention.dense/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3207">
+				<port id="2" precision="FP32" names="1920,1922,input.53">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4791" name="Constant_276292" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4792" name="Constant_276295" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4793" name="Constant_276298" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4794" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2355" name="__module.model.gpt_neox.layers.13/aten::add/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3169">
+				<port id="2" precision="FP32" names="1940">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>60</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4795" name="__module.model.gpt_neox.layers.19.attention/aten::cat/Concat_1" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="2356" name="__module.model.gpt_neox.layers.13/aten::add/Add_1" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>60</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3225,3235,query.117,query.119">
+				<port id="2" precision="FP32" names="1941,1943,hidden_states.59">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4796" name="Constant_276304" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4797" name="Constant_276307" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4798" name="Constant_276310" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2357" name="__module.model.gpt_neox.layers.14.input_layernorm/aten::layer_norm/Multiply" type="Const" version="opset1">
+			<data element_type="i32" shape="1" offset="128867668" size="4" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="I32">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4799" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2358" name="__module.model.gpt_neox.layers.14.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
+			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="I32">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3170,k.39">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2359" name="Constant_37400" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1232345680" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4800" name="__module.model.gpt_neox.layers.19.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
+		<layer id="2360" name="__module.model.gpt_neox.layers.14.post_attention_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3208">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4801" name="Constant_276314" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4802" name="Constant_276313" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2361" name="Constant_37401" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1232355920" size="10240" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4803" name="__module.model.gpt_neox.layers.19.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="2362" name="__module.model.gpt_neox.layers.14.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="2041">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4804" name="__module.model.gpt_neox.layers.19.attention/aten::size/Constant_7" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2363" name="self.model.gpt_neox.layers.14.mlp.dense_h_to_4h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 2560" offset="1232366160" size="26214400" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4805" name="__module.model.gpt_neox.layers.19.attention/aten::size/Gather_7" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="2364" name="Convert_74464" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="3209,3214" />
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4806" name="__module.model.gpt_neox.layers.19.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
+		<layer id="2365" name="self.model.gpt_neox.layers.14.mlp.dense_h_to_4h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 1" offset="1258580560" size="10240" />
 			<output>
-				<port id="2" precision="I64" />
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4807" name="__module.model.gpt_neox.layers.19.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1">
+		<layer id="2366" name="Convert_74467" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64" />
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64" names="3211,3212,3216,3217" />
-			</output>
-		</layer>
-		<layer id="4808" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4809" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1">
+		<layer id="2367" name="self.model.gpt_neox.layers.14.mlp.dense_h_to_4h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4810" name="Constant_276312" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2368" name="self.model.gpt_neox.layers.14.mlp.dense_h_to_4h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="10240, 1" offset="1258590800" size="20480" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4811" name="ScatterUpdate_276315" type="ScatterUpdate" version="opset3">
+		<layer id="2369" name="self.model.gpt_neox.layers.14.mlp.dense_h_to_4h.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4812" name="Constant_276318" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
+		<layer id="2370" name="__module.model.gpt_neox.layers.14.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<rt_info>
-				<attribute name="precise" version="0" />
+				<attribute name="decompression" version="0" />
 			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+			<input>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="4813" name="Constant_276321" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4814" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2371" name="__module.model.gpt_neox.layers.14.mlp.dense_h_to_4h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3218,x2.79">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4815" name="Constant_299267_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
+		<layer id="2372" name="self.model.gpt_neox.layers.14.mlp.dense_h_to_4h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="10240" offset="1258611280" size="20480" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.14.mlp.dense_h_to_4h.bias">
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4816" name="Constant_299267" type="Convert" version="opset1">
+		<layer id="2373" name="__module.model.gpt_neox.layers.14.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
 			</rt_info>
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16">
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4817" name="__module.model.gpt_neox.layers.19.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
+		<layer id="2374" name="__module.model.gpt_neox.layers.14.mlp.dense_h_to_4h/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>10240</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3219">
+				<port id="2" precision="FP32" names="2046">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4818" name="Constant_276366" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2375" name="__module.model.gpt_neox.layers.14.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
+			<data approximation_mode="ERF" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32" names="2047">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4819" name="Constant_276371" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2376" name="self.model.gpt_neox.layers.14.mlp.dense_4h_to_h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 10240" offset="1258631760" size="26214400" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4820" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2377" name="Convert_74474" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+			</input>
+			<output>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+			</output>
+		</layer>
+		<layer id="2378" name="self.model.gpt_neox.layers.14.mlp.dense_4h_to_h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="1284846160" size="2560" />
+			<output>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+			</output>
+		</layer>
+		<layer id="2379" name="Convert_74477" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3213">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4821" name="__module.model.gpt_neox.layers.19.attention/aten::cat/Concat_2" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="2380" name="self.model.gpt_neox.layers.14.mlp.dense_4h_to_h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3221">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2381" name="self.model.gpt_neox.layers.14.mlp.dense_4h_to_h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="1284848720" size="5120" />
+			<output>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4822" name="__module.model.gpt_neox.layers.19.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
+		<layer id="2382" name="self.model.gpt_neox.layers.14.mlp.dense_4h_to_h.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3222">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4823" name="__module.model.gpt_neox.layers.19.attention/aten::add/Add_1" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2383" name="__module.model.gpt_neox.layers.14.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<rt_info>
+				<attribute name="decompression" version="0" />
+			</rt_info>
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2384" name="__module.model.gpt_neox.layers.14.mlp.dense_4h_to_h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>10240</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3223">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4824" name="Constant_276416" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2385" name="self.model.gpt_neox.layers.14.mlp.dense_4h_to_h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="2560" offset="1284853840" size="5120" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.14.mlp.dense_4h_to_h.bias">
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4825" name="Constant_276419" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
+		<layer id="2386" name="__module.model.gpt_neox.layers.14.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<rt_info>
-				<attribute name="precise" version="0" />
+				<attribute name="decompression" version="0" />
 			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+			<input>
+				<port id="0" precision="BF16">
+					<dim>2560</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="4826" name="Constant_276422" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4827" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2387" name="__module.model.gpt_neox.layers.14.mlp.dense_4h_to_h/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3171">
+				<port id="2" precision="FP32" names="2050,input.59">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>60</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4828" name="__module.model.gpt_neox.layers.19.attention/aten::cat/Concat_3" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="2388" name="Constant_37402" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1284858960" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2389" name="__module.model.gpt_neox.layers.14.input_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3227">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4829" name="__module.model.gpt_neox.layers.19.attention/aten::to/Convert_1" type="Concat" version="opset1">
-			<data axis="-2" />
+		<layer id="2390" name="Constant_37403" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1284869200" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2391" name="__module.model.gpt_neox.layers.14.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="present.19.key">
+				<port id="2" precision="FP32" names="1953,hidden_states.57">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4830" name="Constant_337936_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
+		<layer id="2392" name="self.model.gpt_neox.layers.14.attention.query_key_value.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 2560" offset="1284879440" size="19660800" />
 			<output>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4831" name="Constant_337936" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="2393" name="Convert_74978" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" />
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4832" name="Constant_337929" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
+		<layer id="2394" name="self.model.gpt_neox.layers.14.attention.query_key_value.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 1" offset="1304540240" size="7680" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="U8">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4833" name="Constant_337931" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
+		<layer id="2395" name="Convert_74981" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>1</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4834" name="Constant_337933" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
+		<layer id="2396" name="self.model.gpt_neox.layers.14.attention.query_key_value.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
+					<dim>1</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
+				<port id="2" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2397" name="self.model.gpt_neox.layers.14.attention.query_key_value.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="7680, 1" offset="1304547920" size="15360" />
+			<output>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4835" name="Concat_337934" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="2398" name="self.model.gpt_neox.layers.14.attention.query_key_value.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
+			</output>
+		</layer>
+		<layer id="2399" name="self.model.gpt_neox.layers.14.attention.query_key_value.weight/fq_weights_1/convert" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4836" name="Broadcast_337937" type="Broadcast" version="opset3">
-			<data mode="numpy" />
+		<layer id="2400" name="__module.model.gpt_neox.layers.14.attention.query_key_value/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4837" name="ReadValue_330471" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.19.valuepresent.19.value" variable_type="f32" variable_shape="?,32,?,80" />
+		<layer id="2401" name="Constant_37404" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 7680" offset="1304563280" size="30720" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>7680</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2402" name="__module.model.gpt_neox.layers.14.attention.query_key_value/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>7680</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>7680</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="past_key_values.19.value">
+				<port id="2" precision="FP32" names="1960,qkv.57">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4838" name="Constant_323502" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
+		<layer id="2403" name="__module.model.gpt_neox.layers.14.attention/prim::ListConstruct" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201115992" size="32" />
 			<output>
-				<port id="0" precision="I64" />
+				<port id="0" precision="I64">
+					<dim>4</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4839" name="Gather_323503" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="2404" name="__module.model.gpt_neox.layers.14.attention/aten::view/Reshape" type="Reshape" version="opset1">
+			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>7680</dim>
 				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
-				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="3" precision="FP32">
+				<port id="2" precision="FP32" names="1964,qkv.59">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4840" name="Constant_276428" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2405" name="Constant_9536" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4841" name="Constant_276431" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2406" name="Constant_9538" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4842" name="Constant_276434" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2407" name="Constant_9540" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4843" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2408" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Slice" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -74031,17 +36353,20 @@
 					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3165">
+				<port id="5" precision="FP32" names="1965">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -74049,15 +36374,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4844" name="Constant_11290" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="2409" name="Constant_9549" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64" names="3166">
+				<port id="0" precision="I64" names="1966">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4845" name="__module.model.gpt_neox.layers.19.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
+		<layer id="2410" name="__module.model.gpt_neox.layers.14.attention/aten::permute/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -74070,32 +36395,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3167">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4846" name="__module.model.gpt_neox.layers.19.attention/aten::cat/Concat_5" type="Concat" version="opset1">
-			<data axis="-2" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="present.19.value">
+				<port id="2" precision="FP32" names="1967,query.29">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -74103,41 +36403,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4847" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Slice_11" type="Const" version="opset1">
-			<data element_type="boolean" shape="1, 1, 2048, 2048" offset="201244276" size="4194304" />
+		<layer id="2411" name="Constant_9590" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="BOOL" names="3244">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4848" name="Constant_276438" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2412" name="Constant_9592" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>3</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4849" name="Constant_276437" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162298" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2413" name="Constant_9594" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4850" name="__module.model.gpt_neox.layers.19.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="2414" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Slice_1" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -74145,118 +36435,84 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-			</input>
-			<output>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="4851" name="__module.model.gpt_neox.layers.19.attention/aten::size/Constant_8" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="4852" name="__module.model.gpt_neox.layers.19.attention/aten::size/Gather_8" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="3239" />
+				<port id="5" precision="FP32" names="1974,q.29">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4853" name="__module.model.gpt_neox.layers.19.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="2415" name="__module.model.gpt_neox.layers.14.attention/aten::mul/Multiply" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="1980">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4854" name="__module.model.gpt_neox.layers.19.attention/aten::size/Constant_9" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2416" name="Constant_9748" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="4855" name="__module.model.gpt_neox.layers.19.attention/aten::size/Gather_9" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="3237" />
-			</output>
-		</layer>
-		<layer id="4856" name="__module.model.gpt_neox.layers.19.attention/aten::sub/Subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" names="3241,3242" />
-			</output>
-		</layer>
-		<layer id="4857" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Constant_4" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
 			</output>
 		</layer>
-		<layer id="4858" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="2417" name="Constant_9750" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4859" name="Constant_276436" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2418" name="Constant_9752" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4860" name="ScatterUpdate_276439" type="ScatterUpdate" version="opset3">
+		<layer id="2419" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Slice_2" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -74264,146 +36520,89 @@
 				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
 				<port id="4" precision="I64">
-					<dim>3</dim>
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="4861" name="Constant_276440" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="5" precision="FP32" names="1990,x2.57">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4862" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="2420" name="Constant_37405" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4863" name="ScatterUpdate_276441" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
 					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
 					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4864" name="Constant_276444" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438612" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
 			</output>
 		</layer>
-		<layer id="4865" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2421" name="__module.model.gpt_neox.layers.14.attention/aten::neg/Multiply" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="BOOL">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+				<port id="1" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>3</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="BOOL" names="3245">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="4866" name="Constant_276517" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="1991">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4867" name="Constant_276518" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2422" name="Constant_9645" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4868" name="Constant_276514" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2423" name="Constant_9689" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4869" name="Constant_276513" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2424" name="Constant_9691" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4870" name="ScatterUpdate_276519" type="ScatterUpdate" version="opset3">
+		<layer id="2425" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Slice_3" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -74411,183 +36610,128 @@
 				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
 				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4871" name="Constant_276522" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4872" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="BOOL" names="3246,3247">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="5" precision="FP32" names="1985">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4873" name="__module.model.gpt_neox.layers.19.attention/aten::where/Convert_1_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="4874" name="__module.model.gpt_neox.layers.19.attention/aten::where/Convert_1" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="4875" name="__module.model.gpt_neox.layers.19.attention/aten::full/Convert" type="Const" version="opset1">
-			<data element_type="f32" shape="" offset="205438636" size="4" />
-			<output>
-				<port id="0" precision="FP32" names="3234" />
-			</output>
-		</layer>
-		<layer id="4876" name="__module.model.gpt_neox.layers.19.attention/aten::where/Select" type="Select" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2426" name="__module.model.gpt_neox.layers.14.attention/aten::cat/Concat" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
-				<port id="1" precision="FP32" />
-				<port id="2" precision="FP32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="3248,causal_mask.77">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="1993">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4877" name="Add_290452" type="Add" version="opset1">
+		<layer id="2427" name="__module.model.gpt_neox.layers.14.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>1</dim>
 					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="1994">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4878" name="ShapeOf_290453" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="2428" name="__module.model.gpt_neox.layers.14.attention/aten::add/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="1995">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4879" name="Constant_254614" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2429" name="Constant_9600" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4880" name="__module.model.gpt_neox.layers.19.attention/aten::size/Constant_10" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2430" name="Constant_9602" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="4881" name="__module.model.gpt_neox.layers.19.attention/aten::size/Gather_10" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I32" />
-			</input>
+			</output>
+		</layer>
+		<layer id="2431" name="Constant_9604" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="3" precision="I64" names="3232">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4882" name="__module.model.gpt_neox.layers.19.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="2432" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Slice_4" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -74598,93 +36742,154 @@
 				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64" names="3249">
-					<dim>4</dim>
+				<port id="5" precision="FP32" names="1975">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4883" name="__module.model.gpt_neox.layers.19.attention/aten::expand/Abs" type="Abs" version="opset1">
+		<layer id="2433" name="__module.model.gpt_neox.layers.14.attention/aten::cat/Concat_1" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="2013">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4884" name="Maximum_290454" type="Maximum" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2434" name="Constant_52887" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
+			<output>
+				<port id="0" precision="FP32" />
+			</output>
+		</layer>
+		<layer id="2435" name="Constant_52880" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2436" name="Constant_52882" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2437" name="Constant_52884" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2438" name="Concat_52885" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
+				<port id="4" precision="I64">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4885" name="__module.model.gpt_neox.layers.19.attention/aten::add/Add_2" type="Broadcast" version="opset3">
+		<layer id="2439" name="Broadcast_52888" type="Broadcast" version="opset3">
 			<data mode="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
+				<port id="0" precision="FP32" />
 				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3251">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4886" name="__module.model.gpt_neox.layers.19.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
-			<data causal="false" />
+		<layer id="2440" name="ReadValue_50207" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.14.keypresent.14.key" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>-1</dim>
+					<dim>0</dim>
 					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
+			</input>
+			<output>
+				<port id="1" precision="FP32" names="past_key_values.14.key">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-				<port id="2" precision="FP32">
+			</output>
+		</layer>
+		<layer id="2441" name="Constant_47862" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64" />
+			</output>
+		</layer>
+		<layer id="2442" name="Gather_47863" type="Gather" version="opset8">
+			<data batch_dims="0" />
+			<input>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
+				<port id="1" precision="I32">
 					<dim>-1</dim>
 				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3252,3253,sdpa_result.39,tensor.77">
+				<port id="3" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -74692,28 +36897,53 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4887" name="Constant_11731" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="2443" name="Constant_9554" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
-				<port id="0" precision="I64" names="3254">
-					<dim>4</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2444" name="Constant_9556" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2445" name="Constant_9558" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4888" name="__module.model.gpt_neox.layers.19.attention/aten::permute/Transpose_3" type="Transpose" version="opset1">
+		<layer id="2446" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Slice_5" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3255">
+				<port id="5" precision="FP32" names="1968">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -74721,19 +36951,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="4889" name="__module.model.gpt_neox.layers.19.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438650" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2447" name="Constant_9567" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="0" precision="I64" names="1969">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4890" name="__module.model.gpt_neox.layers.19.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
-			<data special_zero="true" />
+		<layer id="2448" name="__module.model.gpt_neox.layers.14.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -74742,986 +36968,960 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>3</dim>
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3260">
+				<port id="2" precision="FP32" names="1970,key.57">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4891" name="self.model.gpt_neox.layers.19.attention.dense.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 2560" offset="1702237394" size="6553600" />
+		<layer id="2449" name="Constant_9610" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4892" name="Convert_22194" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="2450" name="Constant_9612" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4893" name="self.model.gpt_neox.layers.19.attention.dense.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="1708790994" size="2560" />
+		<layer id="2451" name="Constant_9614" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4894" name="Convert_22197" type="Convert" version="opset1">
-			<data destination_type="f16" />
+		<layer id="2452" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Slice_6" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
+				<port id="5" precision="FP32" names="1976,k.29">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4895" name="self.model.gpt_neox.layers.19.attention.dense.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="2453" name="__module.model.gpt_neox.layers.14.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
 					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32" names="1996">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4896" name="self.model.gpt_neox.layers.19.attention.dense.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="1708793554" size="5120" />
+		<layer id="2454" name="Constant_9891" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4897" name="self.model.gpt_neox.layers.19.attention.dense.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+		<layer id="2455" name="Constant_9893" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="2456" name="Constant_9895" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4898" name="self.model.gpt_neox.layers.19.attention.dense.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="2457" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Slice_7" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="5" precision="FP32" names="2006,x2.59">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4899" name="__module.model.gpt_neox.layers.19.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
+		<layer id="2458" name="Constant_37406" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2459" name="__module.model.gpt_neox.layers.14.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>2560</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="2007">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4900" name="Constant_299268_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1708798674" size="5120" />
+		<layer id="2460" name="Constant_9788" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4901" name="Constant_299268" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+		<layer id="2461" name="Constant_9832" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="2462" name="Constant_9834" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4902" name="__module.model.gpt_neox.layers.19.attention.dense/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2463" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Slice_8" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="FP32">
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3263,3265,input.79">
+				<port id="5" precision="FP32" names="2001">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4903" name="__module.model.gpt_neox.layers.19/aten::add/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2464" name="__module.model.gpt_neox.layers.14.attention/aten::cat/Concat_2" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3283">
+				<port id="2" precision="FP32" names="2009">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4904" name="__module.model.gpt_neox.layers.19/aten::add/Add_1" type="Add" version="opset1">
+		<layer id="2465" name="__module.model.gpt_neox.layers.14.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3284,3286,hidden_states.41">
+				<port id="2" precision="FP32" names="2010">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4905" name="__module.model.gpt_neox.layers.20.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867670" size="4" />
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4906" name="__module.model.gpt_neox.layers.20.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
-			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
+		<layer id="2466" name="__module.model.gpt_neox.layers.14.attention/aten::add/Add_1" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="I32">
-					<dim>1</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="2011">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4907" name="Constant_299269_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1708803794" size="5120" />
+		<layer id="2467" name="Constant_9620" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4908" name="Constant_299269" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+		<layer id="2468" name="Constant_9622" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="2469" name="Constant_9624" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4909" name="__module.model.gpt_neox.layers.20.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2470" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Slice_9" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="1" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="4910" name="Constant_299270_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1708808914" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="2" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="4911" name="Constant_299270" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
+				<port id="3" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+				<port id="5" precision="FP32" names="1977">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4912" name="__module.model.gpt_neox.layers.20.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2471" name="__module.model.gpt_neox.layers.14.attention/aten::cat/Concat_3" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3421">
+				<port id="2" precision="FP32" names="2015">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4913" name="self.model.gpt_neox.layers.20.mlp.dense_h_to_4h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 2560" offset="1708814034" size="26214400" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4914" name="Convert_22244" type="Convert" version="opset1">
-			<data destination_type="f16" />
+		<layer id="2472" name="__module.model.gpt_neox.layers.14.attention/aten::cat/Concat_4" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32" names="present.14.key">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4915" name="self.model.gpt_neox.layers.20.mlp.dense_h_to_4h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 1" offset="1735028434" size="10240" />
+		<layer id="2473" name="Constant_52899" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
+				<port id="0" precision="FP32" />
 			</output>
 		</layer>
-		<layer id="4916" name="Convert_22247" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</input>
+		<layer id="2474" name="Constant_52892" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4917" name="self.model.gpt_neox.layers.20.mlp.dense_h_to_4h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</input>
+		<layer id="2475" name="Constant_52894" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="2" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4918" name="self.model.gpt_neox.layers.20.mlp.dense_h_to_4h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="10240, 1" offset="1735038674" size="20480" />
+		<layer id="2476" name="Constant_52896" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4919" name="self.model.gpt_neox.layers.20.mlp.dense_h_to_4h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2477" name="Concat_52897" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
+				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="2" precision="I64">
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="4920" name="self.model.gpt_neox.layers.20.mlp.dense_h_to_4h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="3" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="4" precision="I64">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4921" name="__module.model.gpt_neox.layers.20.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
+		<layer id="2478" name="Broadcast_52900" type="Broadcast" version="opset3">
+			<data mode="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
 				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4922" name="Constant_299271_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 10240" offset="1735059154" size="20480" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4923" name="Constant_299271" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4924" name="__module.model.gpt_neox.layers.20.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2479" name="ReadValue_50209" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.14.valuepresent.14.value" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3426">
+				<port id="1" precision="FP32" names="past_key_values.14.value">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4925" name="__module.model.gpt_neox.layers.20.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
-			<data approximation_mode="ERF" />
+		<layer id="2480" name="Constant_47865" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64" />
+			</output>
+		</layer>
+		<layer id="2481" name="Gather_47866" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>80</dim>
 				</port>
+				<port id="1" precision="I32">
+					<dim>-1</dim>
+				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="3427">
+				<port id="3" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4926" name="self.model.gpt_neox.layers.20.mlp.dense_4h_to_h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 10240" offset="1735079634" size="26214400" />
+		<layer id="2482" name="Constant_9572" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4927" name="Convert_22254" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
+		<layer id="2483" name="Constant_9574" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4928" name="self.model.gpt_neox.layers.20.mlp.dense_4h_to_h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="1761294034" size="2560" />
+		<layer id="2484" name="Constant_9576" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4929" name="Convert_22257" type="Convert" version="opset1">
-			<data destination_type="f16" />
+		<layer id="2485" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Slice_10" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>240</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="4930" name="self.model.gpt_neox.layers.20.mlp.dense_4h_to_h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="3" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="5" precision="FP32" names="1971">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4931" name="self.model.gpt_neox.layers.20.mlp.dense_4h_to_h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="1761296594" size="5120" />
+		<layer id="2486" name="Constant_9585" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
+				<port id="0" precision="I64" names="1972">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4932" name="self.model.gpt_neox.layers.20.mlp.dense_4h_to_h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2487" name="__module.model.gpt_neox.layers.14.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="2" precision="FP32" names="1973">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4933" name="self.model.gpt_neox.layers.20.mlp.dense_4h_to_h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="2488" name="__module.model.gpt_neox.layers.14.attention/aten::cat/Concat_5" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="2" precision="FP32" names="present.14.value">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4934" name="__module.model.gpt_neox.layers.20.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
+		<layer id="2489" name="Constant_9967" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2490" name="ShapeOf_39706" type="ShapeOf" version="opset3">
+			<data output_type="i64" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4935" name="Constant_299272_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1761301714" size="5120" />
+		<layer id="2491" name="Constant_39707" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116168" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4936" name="Constant_299272" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="2492" name="Constant_39708" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64" />
+			</output>
+		</layer>
+		<layer id="2493" name="Gather_39709" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
+					<dim>4</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="3" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4937" name="__module.model.gpt_neox.layers.20.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1">
+		<layer id="2494" name="Add_39712" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3430,input.85">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
+				<port id="2" precision="I64" names="2020">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4938" name="Constant_299273_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1761306834" size="5120" />
+		<layer id="2495" name="Constant_9970" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4939" name="Constant_299273" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="2496" name="__module.model.gpt_neox.layers.14.attention/aten::slice/Slice_14" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
 					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="5" precision="FP32" names="2024">
+					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4940" name="__module.model.gpt_neox.layers.20.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2497" name="__module.model.gpt_neox.layers.14.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+			<data causal="false" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="3" precision="FP32">
+					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="4" precision="FP32" names="2025,attn_output.85">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4941" name="Constant_299274_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1761311954" size="5120" />
+		<layer id="2498" name="__module.model.gpt_neox.layers.14.attention/aten::transpose/Constant" type="Const" version="opset1">
+			<data element_type="i32" shape="4" offset="201116200" size="16" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I32">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4942" name="Constant_299274" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="2499" name="__module.model.gpt_neox.layers.14.attention/aten::transpose/Transpose" type="Transpose" version="opset1">
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I32">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32" names="2026">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4943" name="__module.model.gpt_neox.layers.20.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2500" name="Constant_40998" type="Const" version="opset1">
+			<data element_type="i64" shape="3" offset="201116216" size="24" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>3</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2501" name="__module.model.gpt_neox.layers.14.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
+			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+				<port id="1" precision="I64">
+					<dim>3</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3296">
+				<port id="2" precision="FP32" names="2029">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4944" name="self.model.gpt_neox.layers.20.attention.query_key_value.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 2560" offset="1761317074" size="19660800" />
+		<layer id="2502" name="self.model.gpt_neox.layers.14.attention.dense.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 2560" offset="1304594000" size="6553600" />
 			<output>
 				<port id="0" precision="U8">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4945" name="Convert_22224" type="Convert" version="opset1">
+		<layer id="2503" name="Convert_75330" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP16">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4946" name="self.model.gpt_neox.layers.20.attention.query_key_value.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 1" offset="1780977874" size="7680" />
+		<layer id="2504" name="self.model.gpt_neox.layers.14.attention.dense.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="1311147600" size="2560" />
 			<output>
 				<port id="0" precision="U8">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4947" name="Convert_22227" type="Convert" version="opset1">
+		<layer id="2505" name="Convert_75333" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP16">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4948" name="self.model.gpt_neox.layers.20.attention.query_key_value.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="2506" name="self.model.gpt_neox.layers.14.attention.dense.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP16">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
 				<port id="2" precision="FP16">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4949" name="self.model.gpt_neox.layers.20.attention.query_key_value.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="7680, 1" offset="1780985554" size="15360" />
+		<layer id="2507" name="self.model.gpt_neox.layers.14.attention.dense.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="1311150160" size="5120" />
 			<output>
 				<port id="0" precision="FP16">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4950" name="self.model.gpt_neox.layers.20.attention.query_key_value.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="2508" name="self.model.gpt_neox.layers.14.attention.dense.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP16">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
 				<port id="2" precision="FP16">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4951" name="self.model.gpt_neox.layers.20.attention.query_key_value.weight" type="Convert" version="opset1">
+		<layer id="2509" name="self.model.gpt_neox.layers.14.attention.dense.weight/fq_weights_1/convert" type="Convert" version="opset1">
 			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
 			<input>
 				<port id="0" precision="FP16">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP32">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4952" name="__module.model.gpt_neox.layers.20.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="2510" name="__module.model.gpt_neox.layers.14.attention.dense/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -75730,7 +37930,7 @@
 					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
@@ -75738,826 +37938,838 @@
 				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4953" name="Constant_299275_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 7680" offset="1781000914" size="15360" />
+		<layer id="2511" name="Constant_37407" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1311155280" size="10240" />
 			<output>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4954" name="Constant_299275" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="2512" name="__module.model.gpt_neox.layers.14.attention.dense/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
+				<port id="2" precision="FP32" names="2032,2034,input.57">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4955" name="__module.model.gpt_neox.layers.20.attention.query_key_value/aten::linear/Add" type="Add" version="opset1">
+		<layer id="2513" name="__module.model.gpt_neox.layers.14/aten::add/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3303,qkv.81">
+				<port id="2" precision="FP32" names="2052">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4956" name="__module.model.gpt_neox.layers.20.attention/prim::ListConstruct/Concat" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080154" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4957" name="__module.model.gpt_neox.layers.20.attention/aten::view/Reshape" type="Reshape" version="opset1">
-			<data special_zero="true" />
+		<layer id="2514" name="__module.model.gpt_neox.layers.14/aten::add/Add_1" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3307,qkv.83">
+				<port id="2" precision="FP32" names="2053,2055,hidden_states.63">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4958" name="Constant_276529" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="4959" name="Constant_276532" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4960" name="Constant_276535" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2515" name="__module.model.gpt_neox.layers.15.input_layernorm/aten::layer_norm/Multiply" type="Const" version="opset1">
+			<data element_type="i32" shape="1" offset="128867668" size="4" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="I32">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4961" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2516" name="__module.model.gpt_neox.layers.15.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
+			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="I32">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3308">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4962" name="Constant_11843" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="2517" name="Constant_37408" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1311165520" size="10240" />
 			<output>
-				<port id="0" precision="I64" names="3309">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4963" name="__module.model.gpt_neox.layers.20.attention/aten::permute/Transpose" type="Transpose" version="opset1">
+		<layer id="2518" name="__module.model.gpt_neox.layers.15.post_attention_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3310,query.121">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4964" name="Constant_276541" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2519" name="Constant_37409" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1311175760" size="10240" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4965" name="Constant_276544" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2520" name="__module.model.gpt_neox.layers.15.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="2153">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4966" name="Constant_276547" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2521" name="self.model.gpt_neox.layers.15.mlp.dense_h_to_4h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 2560" offset="1311186000" size="26214400" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4967" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2522" name="Convert_74484" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3317,q.41">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4968" name="self.model.gpt_neox.layers.20.attention.rotary_emb.cos_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201080346" size="81920" />
+		<layer id="2523" name="self.model.gpt_neox.layers.15.mlp.dense_h_to_4h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 1" offset="1337400400" size="10240" />
 			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.20.attention.rotary_emb.cos_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4969" name="self.model.gpt_neox.layers.20.attention.rotary_emb.cos_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="2524" name="Convert_74487" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4970" name="__module.model.gpt_neox.layers.20.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
+		<layer id="2525" name="self.model.gpt_neox.layers.15.mlp.dense_h_to_4h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4971" name="Constant_276553" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2526" name="self.model.gpt_neox.layers.15.mlp.dense_h_to_4h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="10240, 1" offset="1337410640" size="20480" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4972" name="Constant_276556" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2527" name="self.model.gpt_neox.layers.15.mlp.dense_h_to_4h.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4973" name="Constant_276559" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
+		<layer id="2528" name="__module.model.gpt_neox.layers.15.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<rt_info>
-				<attribute name="precise" version="0" />
+				<attribute name="decompression" version="0" />
 			</rt_info>
+			<input>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4974" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2529" name="__module.model.gpt_neox.layers.15.mlp.dense_h_to_4h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3311">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4975" name="Constant_11857" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="2530" name="self.model.gpt_neox.layers.15.mlp.dense_h_to_4h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="10240" offset="1337431120" size="20480" />
 			<output>
-				<port id="0" precision="I64" names="3312">
-					<dim>4</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.15.mlp.dense_h_to_4h.bias">
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2531" name="__module.model.gpt_neox.layers.15.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<rt_info>
+				<attribute name="decompression" version="0" />
+			</rt_info>
+			<input>
+				<port id="0" precision="BF16">
+					<dim>10240</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4976" name="__module.model.gpt_neox.layers.20.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
+		<layer id="2532" name="__module.model.gpt_neox.layers.15.mlp.dense_h_to_4h/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3313,key.121">
+				<port id="2" precision="FP32" names="2158">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4977" name="__module.model.gpt_neox.layers.20.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="2533" name="__module.model.gpt_neox.layers.15.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
+			<data approximation_mode="ERF" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32" names="2159">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4978" name="__module.model.gpt_neox.layers.20.attention/aten::size/Constant_2" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2534" name="self.model.gpt_neox.layers.15.mlp.dense_4h_to_h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 10240" offset="1337451600" size="26214400" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4979" name="__module.model.gpt_neox.layers.20.attention/aten::size/Gather_2" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="2535" name="Convert_74494" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="3321" />
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4980" name="Constant_337949_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
+		<layer id="2536" name="self.model.gpt_neox.layers.15.mlp.dense_4h_to_h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="1363666000" size="2560" />
 			<output>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4981" name="Constant_337949" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="2537" name="Convert_74497" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="4982" name="Constant_337942" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
-			<output>
-				<port id="0" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4983" name="Constant_337944" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64">
+		<layer id="2538" name="self.model.gpt_neox.layers.15.mlp.dense_4h_to_h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4984" name="Constant_337946" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
+		<layer id="2539" name="self.model.gpt_neox.layers.15.mlp.dense_4h_to_h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="1363668560" size="5120" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4985" name="Concat_337947" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="2540" name="self.model.gpt_neox.layers.15.mlp.dense_4h_to_h.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="3" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4986" name="Broadcast_337950" type="Broadcast" version="opset3">
-			<data mode="numpy" />
+		<layer id="2541" name="__module.model.gpt_neox.layers.15.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<rt_info>
+				<attribute name="decompression" version="0" />
+			</rt_info>
 			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4987" name="ReadValue_330477" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.20.keypresent.20.key" variable_type="f32" variable_shape="?,32,?,80" />
+		<layer id="2542" name="__module.model.gpt_neox.layers.15.mlp.dense_4h_to_h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="past_key_values.20.key">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4988" name="Constant_323505" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
+		<layer id="2543" name="self.model.gpt_neox.layers.15.mlp.dense_4h_to_h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="2560" offset="1363673680" size="5120" />
 			<output>
-				<port id="0" precision="I64" />
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.15.mlp.dense_4h_to_h.bias">
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4989" name="Gather_323506" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="2544" name="__module.model.gpt_neox.layers.15.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<rt_info>
+				<attribute name="decompression" version="0" />
+			</rt_info>
+			<input>
+				<port id="0" precision="BF16">
+					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2545" name="__module.model.gpt_neox.layers.15.mlp.dense_4h_to_h/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
 				</port>
-				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="3" precision="FP32">
+				<port id="2" precision="FP32" names="2162,input.63">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4990" name="__module.model.gpt_neox.layers.20.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="2546" name="Constant_37410" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1363678800" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2547" name="__module.model.gpt_neox.layers.15.input_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4991" name="__module.model.gpt_neox.layers.20.attention/aten::size/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2548" name="Constant_37411" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1363689040" size="10240" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="4992" name="__module.model.gpt_neox.layers.20.attention/aten::size/Gather_3" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="3323" />
 			</output>
 		</layer>
-		<layer id="4993" name="__module.model.gpt_neox.layers.20.attention/aten::add_/Add" type="Add" version="opset1">
+		<layer id="2549" name="__module.model.gpt_neox.layers.15.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64" names="3322,3326,3327,seq_len.81" />
+				<port id="2" precision="FP32" names="2065,hidden_states.61">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4994" name="__module.model.gpt_neox.layers.20.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2550" name="self.model.gpt_neox.layers.15.attention.query_key_value.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 2560" offset="1363699280" size="19660800" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4995" name="__module.model.gpt_neox.layers.20.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1">
+		<layer id="2551" name="Convert_74989" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4996" name="__module.model.gpt_neox.layers.20.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2552" name="self.model.gpt_neox.layers.15.attention.query_key_value.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 1" offset="1383360080" size="7680" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="U8">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4997" name="__module.model.gpt_neox.layers.20.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2553" name="Convert_74992" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
+				<port id="0" precision="U8">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
+			</input>
+			<output>
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I64">
+			</output>
+		</layer>
+		<layer id="2554" name="self.model.gpt_neox.layers.15.attention.query_key_value.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3330,3333,cos.81">
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="2" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="4998" name="Constant_38317" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
+		<layer id="2555" name="self.model.gpt_neox.layers.15.attention.query_key_value.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="7680, 1" offset="1383367760" size="15360" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="4999" name="__module.model.gpt_neox.layers.20.attention/aten::index/Gather" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="2556" name="self.model.gpt_neox.layers.15.attention.query_key_value.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
+					<dim>1</dim>
 				</port>
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="3336">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="2" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5000" name="__module.model.gpt_neox.layers.20.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1">
+		<layer id="2557" name="self.model.gpt_neox.layers.15.attention.query_key_value.weight/fq_weights_1/convert" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3337,cos.83">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="1" precision="FP32">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5001" name="__module.model.gpt_neox.layers.20.attention/aten::mul/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2558" name="__module.model.gpt_neox.layers.15.attention.query_key_value/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3341">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5002" name="Constant_276564" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5003" name="Constant_276563" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2559" name="Constant_37412" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 7680" offset="1383383120" size="30720" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
+					<dim>1</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5004" name="__module.model.gpt_neox.layers.20.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="2560" name="__module.model.gpt_neox.layers.15.attention.query_key_value/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>7680</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>7680</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="2072,qkv.61">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5005" name="__module.model.gpt_neox.layers.20.attention/aten::size/Constant_5" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2561" name="__module.model.gpt_neox.layers.15.attention/prim::ListConstruct" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201115992" size="32" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="5006" name="__module.model.gpt_neox.layers.20.attention/aten::size/Gather_5" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
 				<port id="0" precision="I64">
 					<dim>4</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="3342,3347" />
-			</output>
-		</layer>
-		<layer id="5007" name="__module.model.gpt_neox.layers.20.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" />
 			</output>
 		</layer>
-		<layer id="5008" name="__module.model.gpt_neox.layers.20.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1">
+		<layer id="2562" name="__module.model.gpt_neox.layers.15.attention/aten::view/Reshape" type="Reshape" version="opset1">
+			<data special_zero="true" />
 			<input>
-				<port id="0" precision="I64" />
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>7680</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>4</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64" names="3344,3345,3349,3350" />
+				<port id="2" precision="FP32" names="2076,qkv.63">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>240</dim>
+				</port>
 			</output>
-		</layer>
-		<layer id="5009" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Constant_1" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		</layer>
+		<layer id="2563" name="Constant_10166" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5010" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="2564" name="Constant_10168" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5011" name="Constant_276562" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2565" name="Constant_10170" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5012" name="ScatterUpdate_276565" type="ScatterUpdate" version="opset3">
+		<layer id="2566" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Slice" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -76565,148 +38777,155 @@
 				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
 				<port id="4" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="5013" name="Constant_276568" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="5" precision="FP32" names="2077">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5014" name="Constant_276571" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2567" name="Constant_10179" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="I64" names="2078">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5015" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2568" name="__module.model.gpt_neox.layers.15.attention/aten::permute/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3351,x2.81">
+				<port id="2" precision="FP32" names="2079,query.31">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5016" name="Constant_299276_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
+		<layer id="2569" name="Constant_10220" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2570" name="Constant_10222" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2571" name="Constant_10224" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5017" name="Constant_299276" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="2572" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Slice_1" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="5" precision="FP32" names="2086,q.31">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5018" name="__module.model.gpt_neox.layers.20.attention/aten::neg/Multiply" type="Multiply" version="opset1">
+		<layer id="2573" name="__module.model.gpt_neox.layers.15.attention/aten::mul/Multiply" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
+					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3352">
+				<port id="2" precision="FP32" names="2092">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5019" name="Constant_276616" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2574" name="Constant_10378" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5020" name="Constant_276621" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2575" name="Constant_10380" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5021" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2576" name="Constant_10382" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2577" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Slice_2" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -76715,17 +38934,20 @@
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3346">
+				<port id="5" precision="FP32" names="2102,x2.61">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -76733,8 +38955,19 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5022" name="__module.model.gpt_neox.layers.20.attention/aten::cat/Concat" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="2578" name="Constant_37413" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2579" name="__module.model.gpt_neox.layers.15.attention/aten::neg/Multiply" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -76743,75 +38976,51 @@
 					<dim>10</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3354">
+				<port id="2" precision="FP32" names="2103">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5023" name="self.model.gpt_neox.layers.20.attention.rotary_emb.sin_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201162356" size="81920" />
-			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.20.attention.rotary_emb.sin_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5024" name="self.model.gpt_neox.layers.20.attention.rotary_emb.sin_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</input>
+		<layer id="2580" name="Constant_10275" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5025" name="__module.model.gpt_neox.layers.20.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2581" name="Constant_10319" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5026" name="__module.model.gpt_neox.layers.20.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2582" name="Constant_10321" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5027" name="__module.model.gpt_neox.layers.20.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2583" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Slice_3" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>2048</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
@@ -76823,60 +39032,45 @@
 				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3331,3334,sin.81">
+				<port id="5" precision="FP32" names="2097">
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5028" name="Constant_38320" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="5029" name="__module.model.gpt_neox.layers.20.attention/aten::index/Gather_1" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="2584" name="__module.model.gpt_neox.layers.15.attention/aten::cat/Concat" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="3339">
-					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>10</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="5030" name="__module.model.gpt_neox.layers.20.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
+				<port id="1" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>10</dim>
 				</port>
-				<port id="1" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3340,sin.83">
+				<port id="2" precision="FP32" names="2105">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5031" name="__module.model.gpt_neox.layers.20.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
+		<layer id="2585" name="__module.model.gpt_neox.layers.15.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -76893,7 +39087,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3355">
+				<port id="2" precision="FP32" names="2106">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -76901,7 +39095,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5032" name="__module.model.gpt_neox.layers.20.attention/aten::add/Add" type="Add" version="opset1">
+		<layer id="2586" name="__module.model.gpt_neox.layers.15.attention/aten::add/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -76918,7 +39112,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3356">
+				<port id="2" precision="FP32" names="2107">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -76926,41 +39120,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5033" name="Constant_276667" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2587" name="Constant_10230" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5034" name="Constant_276670" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2588" name="Constant_10232" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5035" name="Constant_276673" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2589" name="Constant_10234" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5036" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2590" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Slice_4" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -76969,17 +39153,20 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3318">
+				<port id="5" precision="FP32" names="2087">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -76987,7 +39174,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5037" name="__module.model.gpt_neox.layers.20.attention/aten::cat/Concat_1" type="Concat" version="opset1">
+		<layer id="2591" name="__module.model.gpt_neox.layers.15.attention/aten::cat/Concat_1" type="Concat" version="opset1">
 			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
@@ -77004,7 +39191,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3374,3384,query.123,query.125">
+				<port id="2" precision="FP32" names="2125">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -77012,205 +39199,237 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5038" name="Constant_276679" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2592" name="Constant_52911" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
+			<output>
+				<port id="0" precision="FP32" />
+			</output>
+		</layer>
+		<layer id="2593" name="Constant_52904" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5039" name="Constant_276682" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2594" name="Constant_52906" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5040" name="Constant_276685" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2595" name="Constant_52908" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5041" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2596" name="Concat_52909" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="4" precision="I64">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2597" name="Broadcast_52912" type="Broadcast" version="opset3">
+			<data mode="numpy" />
+			<input>
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3319,k.41">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5042" name="__module.model.gpt_neox.layers.20.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2598" name="ReadValue_50211" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.15.keypresent.15.key" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP32" names="past_key_values.15.key">
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
+			</output>
+		</layer>
+		<layer id="2599" name="Constant_47868" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64" />
+			</output>
+		</layer>
+		<layer id="2600" name="Gather_47869" type="Gather" version="opset8">
+			<data batch_dims="0" />
+			<input>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I32">
 					<dim>-1</dim>
-					<dim>20</dim>
 				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3357">
+				<port id="3" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5043" name="Constant_276689" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2601" name="Constant_10184" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5044" name="Constant_276688" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2602" name="Constant_10186" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5045" name="__module.model.gpt_neox.layers.20.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="2603" name="Constant_10188" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2604" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Slice_5" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
-			</input>
-			<output>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="5046" name="__module.model.gpt_neox.layers.20.attention/aten::size/Constant_7" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="5047" name="__module.model.gpt_neox.layers.20.attention/aten::size/Gather_7" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="3358,3363" />
+				<port id="5" precision="FP32" names="2080">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5048" name="__module.model.gpt_neox.layers.20.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
+		<layer id="2605" name="Constant_10197" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="2" precision="I64" />
+				<port id="0" precision="I64" names="2081">
+					<dim>4</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5049" name="__module.model.gpt_neox.layers.20.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1">
+		<layer id="2606" name="__module.model.gpt_neox.layers.15.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
 			<input>
-				<port id="0" precision="I64" />
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>4</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64" names="3360,3361,3365,3366" />
+				<port id="2" precision="FP32" names="2082,key.61">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5050" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2607" name="Constant_10240" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5051" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="2608" name="Constant_10242" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5052" name="Constant_276687" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2609" name="Constant_10244" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5053" name="ScatterUpdate_276690" type="ScatterUpdate" version="opset3">
+		<layer id="2610" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Slice_6" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -77218,40 +39437,72 @@
 				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="5" precision="FP32" names="2088,k.31">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5054" name="Constant_276693" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2611" name="__module.model.gpt_neox.layers.15.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="2108">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2612" name="Constant_10521" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5055" name="Constant_276696" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2613" name="Constant_10523" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5056" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2614" name="Constant_10525" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2615" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Slice_7" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -77260,17 +39511,20 @@
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3367,x2.83">
+				<port id="5" precision="FP32" names="2118,x2.63">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -77278,32 +39532,10 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5057" name="Constant_299277_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5058" name="Constant_299277" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
+		<layer id="2616" name="Constant_37414" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>1</dim>
@@ -77311,7 +39543,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5059" name="__module.model.gpt_neox.layers.20.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
+		<layer id="2617" name="__module.model.gpt_neox.layers.15.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -77328,7 +39560,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3368">
+				<port id="2" precision="FP32" names="2119">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -77336,30 +39568,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5060" name="Constant_276741" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2618" name="Constant_10418" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5061" name="Constant_276746" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2619" name="Constant_10462" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2620" name="Constant_10464" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5062" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2621" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Slice_8" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -77368,17 +39601,20 @@
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3362">
+				<port id="5" precision="FP32" names="2113">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -77386,7 +39622,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5063" name="__module.model.gpt_neox.layers.20.attention/aten::cat/Concat_2" type="Concat" version="opset1">
+		<layer id="2622" name="__module.model.gpt_neox.layers.15.attention/aten::cat/Concat_2" type="Concat" version="opset1">
 			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
@@ -77403,7 +39639,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3370">
+				<port id="2" precision="FP32" names="2121">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -77411,7 +39647,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5064" name="__module.model.gpt_neox.layers.20.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
+		<layer id="2623" name="__module.model.gpt_neox.layers.15.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -77428,7 +39664,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3371">
+				<port id="2" precision="FP32" names="2122">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -77436,7 +39672,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5065" name="__module.model.gpt_neox.layers.20.attention/aten::add/Add_1" type="Add" version="opset1">
+		<layer id="2624" name="__module.model.gpt_neox.layers.15.attention/aten::add/Add_1" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -77453,7 +39689,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3372">
+				<port id="2" precision="FP32" names="2123">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -77461,41 +39697,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5066" name="Constant_276791" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2625" name="Constant_10250" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5067" name="Constant_276794" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2626" name="Constant_10252" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5068" name="Constant_276797" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2627" name="Constant_10254" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5069" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2628" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Slice_9" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -77504,17 +39730,20 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3320">
+				<port id="5" precision="FP32" names="2089">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -77522,7 +39751,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5070" name="__module.model.gpt_neox.layers.20.attention/aten::cat/Concat_3" type="Concat" version="opset1">
+		<layer id="2629" name="__module.model.gpt_neox.layers.15.attention/aten::cat/Concat_3" type="Concat" version="opset1">
 			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
@@ -77539,7 +39768,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3376">
+				<port id="2" precision="FP32" names="2127">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -77547,7 +39776,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5071" name="__module.model.gpt_neox.layers.20.attention/aten::to/Convert_1" type="Concat" version="opset1">
+		<layer id="2630" name="__module.model.gpt_neox.layers.15.attention/aten::cat/Concat_4" type="Concat" version="opset1">
 			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
@@ -77564,7 +39793,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="present.20.key">
+				<port id="2" precision="FP32" names="present.15.key">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -77572,49 +39801,37 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5072" name="Constant_337962_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="5073" name="Constant_337962" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
+		<layer id="2631" name="Constant_52923" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="1" precision="FP32" />
+				<port id="0" precision="FP32" />
 			</output>
 		</layer>
-		<layer id="5074" name="Constant_337955" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
+		<layer id="2632" name="Constant_52916" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5075" name="Constant_337957" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
+		<layer id="2633" name="Constant_52918" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5076" name="Constant_337959" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
+		<layer id="2634" name="Constant_52920" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5077" name="Concat_337960" type="Concat" version="opset1">
+		<layer id="2635" name="Concat_52921" type="Concat" version="opset1">
 			<data axis="0" />
 			<input>
 				<port id="0" precision="I64">
@@ -77636,7 +39853,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5078" name="Broadcast_337963" type="Broadcast" version="opset3">
+		<layer id="2636" name="Broadcast_52924" type="Broadcast" version="opset3">
 			<data mode="numpy" />
 			<input>
 				<port id="0" precision="FP32" />
@@ -77653,8 +39870,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5079" name="ReadValue_330479" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.20.valuepresent.20.value" variable_type="f32" variable_shape="?,32,?,80" />
+		<layer id="2637" name="ReadValue_50213" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.15.valuepresent.15.value" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -77664,7 +39881,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="past_key_values.20.value">
+				<port id="1" precision="FP32" names="past_key_values.15.value">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -77672,13 +39889,13 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5080" name="Constant_323508" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
+		<layer id="2638" name="Constant_47871" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="5081" name="Gather_323509" type="Gather" version="opset8">
+		<layer id="2639" name="Gather_47872" type="Gather" version="opset8">
 			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="FP32">
@@ -77701,41 +39918,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5082" name="Constant_276803" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2640" name="Constant_10202" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5083" name="Constant_276806" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2641" name="Constant_10204" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5084" name="Constant_276809" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2642" name="Constant_10206" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5085" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2643" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Slice_10" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -77744,17 +39951,20 @@
 					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3314">
+				<port id="5" precision="FP32" names="2083">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -77762,15 +39972,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5086" name="Constant_11871" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="2644" name="Constant_10215" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64" names="3315">
+				<port id="0" precision="I64" names="2084">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5087" name="__module.model.gpt_neox.layers.20.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
+		<layer id="2645" name="__module.model.gpt_neox.layers.15.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -77783,7 +39993,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3316">
+				<port id="2" precision="FP32" names="2085">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -77791,7 +40001,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5088" name="__module.model.gpt_neox.layers.20.attention/aten::cat/Concat_5" type="Concat" version="opset1">
+		<layer id="2646" name="__module.model.gpt_neox.layers.15.attention/aten::cat/Concat_5" type="Concat" version="opset1">
 			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
@@ -77808,7 +40018,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="present.20.value">
+				<port id="2" precision="FP32" names="present.15.value">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -77816,78 +40026,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5089" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Slice_11" type="Const" version="opset1">
-			<data element_type="boolean" shape="1, 1, 2048, 2048" offset="201244276" size="4194304" />
-			<output>
-				<port id="0" precision="BOOL" names="3393">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5090" name="Constant_276813" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5091" name="Constant_276812" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162298" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2647" name="Constant_10597" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5092" name="__module.model.gpt_neox.layers.20.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5093" name="__module.model.gpt_neox.layers.20.attention/aten::size/Constant_8" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="5094" name="__module.model.gpt_neox.layers.20.attention/aten::size/Gather_8" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="3388" />
-			</output>
-		</layer>
-		<layer id="5095" name="__module.model.gpt_neox.layers.20.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3">
+		<layer id="2648" name="ShapeOf_39777" type="ShapeOf" version="opset3">
 			<data output_type="i64" />
 			<input>
 				<port id="0" precision="FP32">
@@ -77903,348 +40050,62 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5096" name="__module.model.gpt_neox.layers.20.attention/aten::size/Constant_9" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="5097" name="__module.model.gpt_neox.layers.20.attention/aten::size/Gather_9" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="3386" />
-			</output>
-		</layer>
-		<layer id="5098" name="__module.model.gpt_neox.layers.20.attention/aten::sub/Subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" names="3390,3391" />
-			</output>
-		</layer>
-		<layer id="5099" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Constant_4" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="5100" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5101" name="Constant_276811" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2649" name="Constant_39778" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116168" size="8" />
 			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5102" name="ScatterUpdate_276814" type="ScatterUpdate" version="opset3">
-			<input>
 				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
-				</port>
 			</output>
 		</layer>
-		<layer id="5103" name="Constant_276815" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2650" name="Constant_39779" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5104" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1">
-			<input>
 				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
 			</output>
 		</layer>
-		<layer id="5105" name="ScatterUpdate_276816" type="ScatterUpdate" version="opset3">
+		<layer id="2651" name="Gather_39780" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="I64">
-					<dim>3</dim>
+					<dim>4</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5106" name="Constant_276819" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438612" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5107" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>3</dim>
-				</port>
 				<port id="3" precision="I64">
-					<dim>3</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="BOOL" names="3394">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5108" name="Constant_276892" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5109" name="Constant_276893" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5110" name="Constant_276889" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5111" name="Constant_276888" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5112" name="ScatterUpdate_276894" type="ScatterUpdate" version="opset3">
+		<layer id="2652" name="Add_39783" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5113" name="Constant_276897" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5114" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="BOOL">
 					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="BOOL" names="3395,3396">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5115" name="__module.model.gpt_neox.layers.20.attention/aten::where/Convert_1_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="5116" name="__module.model.gpt_neox.layers.20.attention/aten::where/Convert_1" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="5117" name="__module.model.gpt_neox.layers.20.attention/aten::full/Convert" type="Const" version="opset1">
-			<data element_type="f32" shape="" offset="205438636" size="4" />
-			<output>
-				<port id="0" precision="FP32" names="3383" />
-			</output>
-		</layer>
-		<layer id="5118" name="__module.model.gpt_neox.layers.20.attention/aten::where/Select" type="Select" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="BOOL">
 					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
 				</port>
-				<port id="1" precision="FP32" />
-				<port id="2" precision="FP32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="3397,causal_mask.81">
-					<dim>1</dim>
+				<port id="2" precision="I64" names="2132">
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5119" name="Add_290458" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-				</port>
-			</input>
+		<layer id="2653" name="Constant_10600" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5120" name="ShapeOf_290459" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="2654" name="__module.model.gpt_neox.layers.15.attention/aten::slice/Slice_14" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -78252,56 +40113,6 @@
 					<dim>-1</dim>
 					<dim>-1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5121" name="Constant_254629" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5122" name="__module.model.gpt_neox.layers.20.attention/aten::size/Constant_10" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="5123" name="__module.model.gpt_neox.layers.20.attention/aten::size/Gather_10" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="3381">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5124" name="__module.model.gpt_neox.layers.20.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1">
-			<data axis="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
@@ -78311,64 +40122,20 @@
 				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64" names="3398">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5125" name="__module.model.gpt_neox.layers.20.attention/aten::expand/Abs" type="Abs" version="opset1">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5126" name="Maximum_290460" type="Maximum" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5127" name="__module.model.gpt_neox.layers.20.attention/aten::add/Add_2" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3400">
-					<dim>-1</dim>
+				<port id="5" precision="FP32" names="2136">
 					<dim>-1</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
 					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5128" name="__module.model.gpt_neox.layers.20.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+		<layer id="2655" name="__module.model.gpt_neox.layers.15.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
 			<data causal="false" />
 			<input>
 				<port id="0" precision="FP32">
@@ -78391,13 +40158,13 @@
 				</port>
 				<port id="3" precision="FP32">
 					<dim>-1</dim>
-					<dim>-1</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
 					<dim>-1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3401,3402,sdpa_result.41,tensor.81">
+				<port id="4" precision="FP32" names="2137,attn_output.91">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -78405,15 +40172,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5129" name="Constant_12312" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="2656" name="__module.model.gpt_neox.layers.15.attention/aten::transpose/Constant" type="Const" version="opset1">
+			<data element_type="i32" shape="4" offset="201116200" size="16" />
 			<output>
-				<port id="0" precision="I64" names="3403">
+				<port id="0" precision="I32">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5130" name="__module.model.gpt_neox.layers.20.attention/aten::permute/Transpose_3" type="Transpose" version="opset1">
+		<layer id="2657" name="__module.model.gpt_neox.layers.15.attention/aten::transpose/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -78421,12 +40188,12 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="I32">
 					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3404">
+				<port id="2" precision="FP32" names="2138">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -78434,18 +40201,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5131" name="__module.model.gpt_neox.layers.20.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438650" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2658" name="Constant_40999" type="Const" version="opset1">
+			<data element_type="i64" shape="3" offset="201116216" size="24" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>3</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5132" name="__module.model.gpt_neox.layers.20.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
+		<layer id="2659" name="__module.model.gpt_neox.layers.15.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
 			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -78459,15 +40223,15 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3409">
+				<port id="2" precision="FP32" names="2141">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5133" name="self.model.gpt_neox.layers.20.attention.dense.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 2560" offset="1781016274" size="6553600" />
+		<layer id="2660" name="self.model.gpt_neox.layers.15.attention.dense.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 2560" offset="1383413840" size="6553600" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>2560</dim>
@@ -78475,7 +40239,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5134" name="Convert_22234" type="Convert" version="opset1">
+		<layer id="2661" name="Convert_75341" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -78490,8 +40254,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5135" name="self.model.gpt_neox.layers.20.attention.dense.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="1787569874" size="2560" />
+		<layer id="2662" name="self.model.gpt_neox.layers.15.attention.dense.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="1389967440" size="2560" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>2560</dim>
@@ -78499,7 +40263,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5136" name="Convert_22237" type="Convert" version="opset1">
+		<layer id="2663" name="Convert_75344" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -78514,7 +40278,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5137" name="self.model.gpt_neox.layers.20.attention.dense.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="2664" name="self.model.gpt_neox.layers.15.attention.dense.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -78533,8 +40297,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5138" name="self.model.gpt_neox.layers.20.attention.dense.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="1787572434" size="5120" />
+		<layer id="2665" name="self.model.gpt_neox.layers.15.attention.dense.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="1389970000" size="5120" />
 			<output>
 				<port id="0" precision="FP16">
 					<dim>2560</dim>
@@ -78542,7 +40306,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5139" name="self.model.gpt_neox.layers.20.attention.dense.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="2666" name="self.model.gpt_neox.layers.15.attention.dense.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -78561,11 +40325,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5140" name="self.model.gpt_neox.layers.20.attention.dense.weight" type="Convert" version="opset1">
+		<layer id="2667" name="self.model.gpt_neox.layers.15.attention.dense.weight/fq_weights_1/convert" type="Convert" version="opset1">
 			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
 			<input>
 				<port id="0" precision="FP16">
 					<dim>2560</dim>
@@ -78579,7 +40340,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5141" name="__module.model.gpt_neox.layers.20.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="2668" name="__module.model.gpt_neox.layers.15.attention.dense/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -78600,37 +40361,17 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5142" name="Constant_299278_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1787577554" size="5120" />
+		<layer id="2669" name="Constant_37415" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1389975120" size="10240" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5143" name="Constant_299278" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5144" name="__module.model.gpt_neox.layers.20.attention.dense/aten::linear/Add" type="Add" version="opset1">
+		<layer id="2670" name="__module.model.gpt_neox.layers.15.attention.dense/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -78645,14 +40386,14 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3412,3414,input.83">
+				<port id="2" precision="FP32" names="2144,2146,input.61">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5145" name="__module.model.gpt_neox.layers.20/aten::add/Add" type="Add" version="opset1">
+		<layer id="2671" name="__module.model.gpt_neox.layers.15/aten::add/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -78667,14 +40408,14 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3432">
+				<port id="2" precision="FP32" names="2164">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5146" name="__module.model.gpt_neox.layers.20/aten::add/Add_1" type="Add" version="opset1">
+		<layer id="2672" name="__module.model.gpt_neox.layers.15/aten::add/Add_1" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -78689,22 +40430,22 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3433,3435,hidden_states.43">
+				<port id="2" precision="FP32" names="2165,2167,hidden_states.67">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5147" name="__module.model.gpt_neox.layers.21.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867670" size="4" />
+		<layer id="2673" name="__module.model.gpt_neox.layers.16.input_layernorm/aten::layer_norm/Multiply" type="Const" version="opset1">
+			<data element_type="i32" shape="1" offset="128867668" size="4" />
 			<output>
 				<port id="0" precision="I32">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5148" name="__module.model.gpt_neox.layers.21.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
+		<layer id="2674" name="__module.model.gpt_neox.layers.16.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
 			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
 			<input>
 				<port id="0" precision="FP32">
@@ -78724,37 +40465,17 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5149" name="Constant_299279_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1787582674" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5150" name="Constant_299279" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="2675" name="Constant_37416" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1389985360" size="10240" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5151" name="__module.model.gpt_neox.layers.21.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
+		<layer id="2676" name="__module.model.gpt_neox.layers.16.post_attention_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -78776,37 +40497,17 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5152" name="Constant_299280_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1787587794" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5153" name="Constant_299280" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="2677" name="Constant_37417" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1389995600" size="10240" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5154" name="__module.model.gpt_neox.layers.21.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+		<layer id="2678" name="__module.model.gpt_neox.layers.16.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -78821,15 +40522,15 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3570">
+				<port id="2" precision="FP32" names="2265">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5155" name="self.model.gpt_neox.layers.21.mlp.dense_h_to_4h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 2560" offset="1787592914" size="26214400" />
+		<layer id="2679" name="self.model.gpt_neox.layers.16.mlp.dense_h_to_4h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 2560" offset="1390005840" size="26214400" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>10240</dim>
@@ -78837,7 +40538,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5156" name="Convert_22284" type="Convert" version="opset1">
+		<layer id="2680" name="Convert_74504" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -78852,8 +40553,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5157" name="self.model.gpt_neox.layers.21.mlp.dense_h_to_4h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 1" offset="1813807314" size="10240" />
+		<layer id="2681" name="self.model.gpt_neox.layers.16.mlp.dense_h_to_4h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 1" offset="1416220240" size="10240" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>10240</dim>
@@ -78861,7 +40562,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5158" name="Convert_22287" type="Convert" version="opset1">
+		<layer id="2682" name="Convert_74507" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -78876,7 +40577,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5159" name="self.model.gpt_neox.layers.21.mlp.dense_h_to_4h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="2683" name="self.model.gpt_neox.layers.16.mlp.dense_h_to_4h.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -78895,8 +40596,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5160" name="self.model.gpt_neox.layers.21.mlp.dense_h_to_4h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="10240, 1" offset="1813817554" size="20480" />
+		<layer id="2684" name="self.model.gpt_neox.layers.16.mlp.dense_h_to_4h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="10240, 1" offset="1416230480" size="20480" />
 			<output>
 				<port id="0" precision="FP16">
 					<dim>10240</dim>
@@ -78904,7 +40605,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5161" name="self.model.gpt_neox.layers.21.mlp.dense_h_to_4h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="2685" name="self.model.gpt_neox.layers.16.mlp.dense_h_to_4h.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -78923,7 +40624,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5162" name="self.model.gpt_neox.layers.21.mlp.dense_h_to_4h.weight" type="Convert" version="opset1">
+		<layer id="2686" name="__module.model.gpt_neox.layers.16.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
@@ -78941,7 +40642,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5163" name="__module.model.gpt_neox.layers.21.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="2687" name="__module.model.gpt_neox.layers.16.mlp.dense_h_to_4h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -78962,37 +40663,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5164" name="Constant_299281_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 10240" offset="1813838034" size="20480" />
+		<layer id="2688" name="self.model.gpt_neox.layers.16.mlp.dense_h_to_4h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="10240" offset="1416250960" size="20480" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.16.mlp.dense_h_to_4h.bias">
 					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5165" name="Constant_299281" type="Convert" version="opset1">
+		<layer id="2689" name="__module.model.gpt_neox.layers.16.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
 			</rt_info>
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16">
 					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5166" name="__module.model.gpt_neox.layers.21.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1">
+		<layer id="2690" name="__module.model.gpt_neox.layers.16.mlp.dense_h_to_4h/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -79001,20 +40696,18 @@
 					<dim>10240</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3575">
+				<port id="2" precision="FP32" names="2270">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5167" name="__module.model.gpt_neox.layers.21.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
+		<layer id="2691" name="__module.model.gpt_neox.layers.16.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
 			<data approximation_mode="ERF" />
 			<input>
 				<port id="0" precision="FP32">
@@ -79024,15 +40717,15 @@
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="3576">
+				<port id="1" precision="FP32" names="2271">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5168" name="self.model.gpt_neox.layers.21.mlp.dense_4h_to_h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 10240" offset="1813858514" size="26214400" />
+		<layer id="2692" name="self.model.gpt_neox.layers.16.mlp.dense_4h_to_h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 10240" offset="1416271440" size="26214400" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>2560</dim>
@@ -79040,7 +40733,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5169" name="Convert_22294" type="Convert" version="opset1">
+		<layer id="2693" name="Convert_74514" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -79055,8 +40748,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5170" name="self.model.gpt_neox.layers.21.mlp.dense_4h_to_h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="1840072914" size="2560" />
+		<layer id="2694" name="self.model.gpt_neox.layers.16.mlp.dense_4h_to_h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="1442485840" size="2560" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>2560</dim>
@@ -79064,7 +40757,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5171" name="Convert_22297" type="Convert" version="opset1">
+		<layer id="2695" name="Convert_74517" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -79079,7 +40772,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5172" name="self.model.gpt_neox.layers.21.mlp.dense_4h_to_h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="2696" name="self.model.gpt_neox.layers.16.mlp.dense_4h_to_h.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -79098,8 +40791,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5173" name="self.model.gpt_neox.layers.21.mlp.dense_4h_to_h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="1840075474" size="5120" />
+		<layer id="2697" name="self.model.gpt_neox.layers.16.mlp.dense_4h_to_h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="1442488400" size="5120" />
 			<output>
 				<port id="0" precision="FP16">
 					<dim>2560</dim>
@@ -79107,7 +40800,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5174" name="self.model.gpt_neox.layers.21.mlp.dense_4h_to_h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="2698" name="self.model.gpt_neox.layers.16.mlp.dense_4h_to_h.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -79126,7 +40819,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5175" name="self.model.gpt_neox.layers.21.mlp.dense_4h_to_h.weight" type="Convert" version="opset1">
+		<layer id="2699" name="__module.model.gpt_neox.layers.16.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
@@ -79144,7 +40837,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5176" name="__module.model.gpt_neox.layers.21.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="2700" name="__module.model.gpt_neox.layers.16.mlp.dense_4h_to_h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -79165,37 +40858,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5177" name="Constant_299282_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1840080594" size="5120" />
+		<layer id="2701" name="self.model.gpt_neox.layers.16.mlp.dense_4h_to_h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="2560" offset="1442493520" size="5120" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.16.mlp.dense_4h_to_h.bias">
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5178" name="Constant_299282" type="Convert" version="opset1">
+		<layer id="2702" name="__module.model.gpt_neox.layers.16.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
 			</rt_info>
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16">
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5179" name="__module.model.gpt_neox.layers.21.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1">
+		<layer id="2703" name="__module.model.gpt_neox.layers.16.mlp.dense_4h_to_h/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -79204,50 +40891,28 @@
 					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3579,input.89">
+				<port id="2" precision="FP32" names="2274,input.67">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5180" name="Constant_299283_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1840085714" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5181" name="Constant_299283" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="2704" name="Constant_37418" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1442498640" size="10240" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5182" name="__module.model.gpt_neox.layers.21.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
+		<layer id="2705" name="__module.model.gpt_neox.layers.16.input_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -79269,37 +40934,17 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5183" name="Constant_299284_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1840090834" size="5120" />
+		<layer id="2706" name="Constant_37419" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1442508880" size="10240" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5184" name="Constant_299284" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5185" name="__module.model.gpt_neox.layers.21.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+		<layer id="2707" name="__module.model.gpt_neox.layers.16.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -79314,15 +40959,15 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3445">
+				<port id="2" precision="FP32" names="2177,hidden_states.65">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5186" name="self.model.gpt_neox.layers.21.attention.query_key_value.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 2560" offset="1840095954" size="19660800" />
+		<layer id="2708" name="self.model.gpt_neox.layers.16.attention.query_key_value.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 2560" offset="1442519120" size="19660800" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>7680</dim>
@@ -79330,7 +40975,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5187" name="Convert_22264" type="Convert" version="opset1">
+		<layer id="2709" name="Convert_75000" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -79345,8 +40990,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5188" name="self.model.gpt_neox.layers.21.attention.query_key_value.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 1" offset="1859756754" size="7680" />
+		<layer id="2710" name="self.model.gpt_neox.layers.16.attention.query_key_value.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 1" offset="1462179920" size="7680" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>7680</dim>
@@ -79354,7 +40999,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5189" name="Convert_22267" type="Convert" version="opset1">
+		<layer id="2711" name="Convert_75003" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -79369,7 +41014,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5190" name="self.model.gpt_neox.layers.21.attention.query_key_value.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="2712" name="self.model.gpt_neox.layers.16.attention.query_key_value.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -79388,8 +41033,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5191" name="self.model.gpt_neox.layers.21.attention.query_key_value.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="7680, 1" offset="1859764434" size="15360" />
+		<layer id="2713" name="self.model.gpt_neox.layers.16.attention.query_key_value.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="7680, 1" offset="1462187600" size="15360" />
 			<output>
 				<port id="0" precision="FP16">
 					<dim>7680</dim>
@@ -79397,7 +41042,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5192" name="self.model.gpt_neox.layers.21.attention.query_key_value.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="2714" name="self.model.gpt_neox.layers.16.attention.query_key_value.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -79416,11 +41061,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5193" name="self.model.gpt_neox.layers.21.attention.query_key_value.weight" type="Convert" version="opset1">
+		<layer id="2715" name="self.model.gpt_neox.layers.16.attention.query_key_value.weight/fq_weights_1/convert" type="Convert" version="opset1">
 			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
 			<input>
 				<port id="0" precision="FP16">
 					<dim>7680</dim>
@@ -79434,7 +41076,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5194" name="__module.model.gpt_neox.layers.21.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="2716" name="__module.model.gpt_neox.layers.16.attention.query_key_value/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -79455,37 +41097,17 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5195" name="Constant_299285_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 7680" offset="1859779794" size="15360" />
+		<layer id="2717" name="Constant_37420" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 7680" offset="1462202960" size="30720" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5196" name="Constant_299285" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5197" name="__module.model.gpt_neox.layers.21.attention.query_key_value/aten::linear/Add" type="Add" version="opset1">
+		<layer id="2718" name="__module.model.gpt_neox.layers.16.attention.query_key_value/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -79500,25 +41122,22 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3452,qkv.85">
+				<port id="2" precision="FP32" names="2184,qkv.65">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5198" name="__module.model.gpt_neox.layers.21.attention/prim::ListConstruct/Concat" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080154" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2719" name="__module.model.gpt_neox.layers.16.attention/prim::ListConstruct" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201115992" size="32" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5199" name="__module.model.gpt_neox.layers.21.attention/aten::view/Reshape" type="Reshape" version="opset1">
+		<layer id="2720" name="__module.model.gpt_neox.layers.16.attention/aten::view/Reshape" type="Reshape" version="opset1">
 			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -79531,7 +41150,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3456,qkv.87">
+				<port id="2" precision="FP32" names="2188,qkv.67">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -79539,41 +41158,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5200" name="Constant_276904" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2721" name="Constant_10796" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5201" name="Constant_276907" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2722" name="Constant_10798" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5202" name="Constant_276910" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2723" name="Constant_10800" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5203" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2724" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Slice" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -79582,17 +41191,20 @@
 					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3457">
+				<port id="5" precision="FP32" names="2189">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -79600,15 +41212,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5204" name="Constant_12424" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="2725" name="Constant_10809" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64" names="3458">
+				<port id="0" precision="I64" names="2190">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5205" name="__module.model.gpt_neox.layers.21.attention/aten::permute/Transpose" type="Transpose" version="opset1">
+		<layer id="2726" name="__module.model.gpt_neox.layers.16.attention/aten::permute/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -79621,7 +41233,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3459,query.127">
+				<port id="2" precision="FP32" names="2191,query.33">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -79629,41 +41241,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5206" name="Constant_276916" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2727" name="Constant_10850" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5207" name="Constant_276919" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2728" name="Constant_10852" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5208" name="Constant_276922" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2729" name="Constant_10854" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5209" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2730" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Slice_1" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -79672,17 +41274,20 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3466,q.43">
+				<port id="5" precision="FP32" names="2198,q.33">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -79690,219 +41295,152 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5210" name="self.model.gpt_neox.layers.21.attention.rotary_emb.cos_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201080346" size="81920" />
-			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.21.attention.rotary_emb.cos_cached">
-					<dim>2048</dim>
+		<layer id="2731" name="__module.model.gpt_neox.layers.16.attention/aten::mul/Multiply" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="5211" name="self.model.gpt_neox.layers.21.attention.rotary_emb.cos_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
+				<port id="2" precision="FP32" names="2204">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5212" name="__module.model.gpt_neox.layers.21.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2732" name="Constant_11008" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5213" name="Constant_276928" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2733" name="Constant_11010" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5214" name="Constant_276931" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5215" name="Constant_276934" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2734" name="Constant_11012" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5216" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2735" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Slice_2" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>240</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3460">
-					<dim>-1</dim>
+				<port id="5" precision="FP32" names="2214,x2.65">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5217" name="Constant_12438" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="2736" name="Constant_37421" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="0" precision="I64" names="3461">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5218" name="__module.model.gpt_neox.layers.21.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
+		<layer id="2737" name="__module.model.gpt_neox.layers.16.attention/aten::neg/Multiply" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3462,key.127">
+				<port id="2" precision="FP32" names="2215">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5219" name="__module.model.gpt_neox.layers.21.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5220" name="__module.model.gpt_neox.layers.21.attention/aten::size/Constant_2" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="5221" name="__module.model.gpt_neox.layers.21.attention/aten::size/Gather_2" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="3470" />
-			</output>
-		</layer>
-		<layer id="5222" name="Constant_337975_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="5223" name="Constant_337975" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="5224" name="Constant_337968" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
+		<layer id="2738" name="Constant_10905" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5225" name="Constant_337970" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
+		<layer id="2739" name="Constant_10949" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5226" name="Constant_337972" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
+		<layer id="2740" name="Constant_10951" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5227" name="Concat_337973" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="2741" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Slice_3" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -79913,163 +41451,125 @@
 				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
 				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5228" name="Broadcast_337976" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="5" precision="FP32" names="2209">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5229" name="ReadValue_330481" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.21.keypresent.21.key" variable_type="f32" variable_shape="?,32,?,80" />
+		<layer id="2742" name="__module.model.gpt_neox.layers.16.attention/aten::cat/Concat" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="past_key_values.21.key">
+				<port id="2" precision="FP32" names="2217">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5230" name="Constant_323511" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64" />
-			</output>
-		</layer>
-		<layer id="5231" name="Gather_323512" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="2743" name="__module.model.gpt_neox.layers.16.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="I32">
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
-				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="3" precision="FP32">
+				<port id="2" precision="FP32" names="2218">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5232" name="__module.model.gpt_neox.layers.21.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="2744" name="__module.model.gpt_neox.layers.16.attention/aten::add/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="2219">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5233" name="__module.model.gpt_neox.layers.21.attention/aten::size/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2745" name="Constant_10860" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="5234" name="__module.model.gpt_neox.layers.21.attention/aten::size/Gather_3" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="3472" />
-			</output>
-		</layer>
-		<layer id="5235" name="__module.model.gpt_neox.layers.21.attention/aten::add_/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" names="3471,3475,3476,seq_len.85" />
-			</output>
-		</layer>
-		<layer id="5236" name="__module.model.gpt_neox.layers.21.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
 			</output>
 		</layer>
-		<layer id="5237" name="__module.model.gpt_neox.layers.21.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="2746" name="Constant_10862" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5238" name="__module.model.gpt_neox.layers.21.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2747" name="Constant_10864" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5239" name="__module.model.gpt_neox.layers.21.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2748" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Slice_4" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -80080,197 +41580,192 @@
 				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3479,3482,cos.85">
+				<port id="5" precision="FP32" names="2199">
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5240" name="Constant_38323" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="5241" name="__module.model.gpt_neox.layers.21.attention/aten::index/Gather" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="2749" name="__module.model.gpt_neox.layers.16.attention/aten::cat/Concat_1" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
-				<port id="1" precision="I32">
+				<port id="1" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="3485">
+				<port id="2" precision="FP32" names="2237">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5242" name="__module.model.gpt_neox.layers.21.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64" />
-			</input>
+		<layer id="2750" name="Constant_52935" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="2" precision="FP32" names="3486,cos.87">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
+				<port id="0" precision="FP32" />
 			</output>
 		</layer>
-		<layer id="5243" name="__module.model.gpt_neox.layers.21.attention/aten::mul/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
+		<layer id="2751" name="Constant_52928" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
-				<port id="2" precision="FP32" names="3490">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5244" name="Constant_276939" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2752" name="Constant_52930" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5245" name="Constant_276938" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2753" name="Constant_52932" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5246" name="__module.model.gpt_neox.layers.21.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="2754" name="Concat_52933" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
+				<port id="4" precision="I64">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5247" name="__module.model.gpt_neox.layers.21.attention/aten::size/Constant_5" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="5248" name="__module.model.gpt_neox.layers.21.attention/aten::size/Gather_5" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="2755" name="Broadcast_52936" type="Broadcast" version="opset3">
+			<data mode="numpy" />
 			<input>
-				<port id="0" precision="I64">
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="3491,3496" />
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5249" name="__module.model.gpt_neox.layers.21.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
+		<layer id="2756" name="ReadValue_50215" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.16.keypresent.16.key" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64" />
+				<port id="1" precision="FP32" names="past_key_values.16.key">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2757" name="Constant_47874" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="5250" name="__module.model.gpt_neox.layers.21.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1">
+		<layer id="2758" name="Gather_47875" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
-				<port id="0" precision="I64" />
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I32">
+					<dim>-1</dim>
+				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="1" precision="I64" names="3493,3494,3498,3499" />
+				<port id="3" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5251" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Constant_1" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2759" name="Constant_10814" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5252" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="2760" name="Constant_10816" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5253" name="Constant_276937" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2761" name="Constant_10818" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5254" name="ScatterUpdate_276940" type="ScatterUpdate" version="opset3">
+		<layer id="2762" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Slice_5" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -80278,148 +41773,155 @@
 				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
 				<port id="4" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="5255" name="Constant_276943" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="5" precision="FP32" names="2192">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5256" name="Constant_276946" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2763" name="Constant_10827" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="I64" names="2193">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5257" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2764" name="__module.model.gpt_neox.layers.16.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3500,x2.85">
+				<port id="2" precision="FP32" names="2194,key.65">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5258" name="Constant_299286_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
+		<layer id="2765" name="Constant_10870" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2766" name="Constant_10872" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2767" name="Constant_10874" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5259" name="Constant_299286" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="2768" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Slice_6" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="5" precision="FP32" names="2200,k.33">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5260" name="__module.model.gpt_neox.layers.21.attention/aten::neg/Multiply" type="Multiply" version="opset1">
+		<layer id="2769" name="__module.model.gpt_neox.layers.16.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
+					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3501">
+				<port id="2" precision="FP32" names="2220">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5261" name="Constant_276991" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2770" name="Constant_11151" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5262" name="Constant_276996" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2771" name="Constant_11153" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2772" name="Constant_11155" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5263" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2773" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Slice_7" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -80428,17 +41930,20 @@
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3495">
+				<port id="5" precision="FP32" names="2230,x2.67">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -80446,8 +41951,19 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5264" name="__module.model.gpt_neox.layers.21.attention/aten::cat/Concat" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="2774" name="Constant_37422" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2775" name="__module.model.gpt_neox.layers.16.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -80456,75 +41972,51 @@
 					<dim>10</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3503">
+				<port id="2" precision="FP32" names="2231">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5265" name="self.model.gpt_neox.layers.21.attention.rotary_emb.sin_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201162356" size="81920" />
-			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.21.attention.rotary_emb.sin_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5266" name="self.model.gpt_neox.layers.21.attention.rotary_emb.sin_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</input>
+		<layer id="2776" name="Constant_11048" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5267" name="__module.model.gpt_neox.layers.21.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2777" name="Constant_11092" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5268" name="__module.model.gpt_neox.layers.21.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2778" name="Constant_11094" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5269" name="__module.model.gpt_neox.layers.21.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2779" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Slice_8" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>2048</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
@@ -80536,60 +42028,45 @@
 				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3480,3483,sin.85">
+				<port id="5" precision="FP32" names="2225">
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5270" name="Constant_38326" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="5271" name="__module.model.gpt_neox.layers.21.attention/aten::index/Gather_1" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="2780" name="__module.model.gpt_neox.layers.16.attention/aten::cat/Concat_2" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="3488">
-					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>10</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="5272" name="__module.model.gpt_neox.layers.21.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
+				<port id="1" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>10</dim>
 				</port>
-				<port id="1" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3489,sin.87">
+				<port id="2" precision="FP32" names="2233">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5273" name="__module.model.gpt_neox.layers.21.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
+		<layer id="2781" name="__module.model.gpt_neox.layers.16.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -80606,7 +42083,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3504">
+				<port id="2" precision="FP32" names="2234">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -80614,7 +42091,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5274" name="__module.model.gpt_neox.layers.21.attention/aten::add/Add" type="Add" version="opset1">
+		<layer id="2782" name="__module.model.gpt_neox.layers.16.attention/aten::add/Add_1" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -80631,7 +42108,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3505">
+				<port id="2" precision="FP32" names="2235">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -80639,41 +42116,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5275" name="Constant_277042" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2783" name="Constant_10880" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5276" name="Constant_277045" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2784" name="Constant_10882" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5277" name="Constant_277048" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2785" name="Constant_10884" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5278" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2786" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Slice_9" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -80682,17 +42149,20 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3467">
+				<port id="5" precision="FP32" names="2201">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -80700,7 +42170,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5279" name="__module.model.gpt_neox.layers.21.attention/aten::cat/Concat_1" type="Concat" version="opset1">
+		<layer id="2787" name="__module.model.gpt_neox.layers.16.attention/aten::cat/Concat_3" type="Concat" version="opset1">
 			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
@@ -80717,7 +42187,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3523,3533,query.129,query.131">
+				<port id="2" precision="FP32" names="2239">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -80725,41 +42195,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5280" name="Constant_277054" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5281" name="Constant_277057" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5282" name="Constant_277060" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5283" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2788" name="__module.model.gpt_neox.layers.16.attention/aten::cat/Concat_4" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -80767,163 +42204,57 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="3468,k.43">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5284" name="__module.model.gpt_neox.layers.21.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3506">
+				<port id="2" precision="FP32" names="present.16.key">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5285" name="Constant_277064" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2789" name="Constant_52947" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
+				<port id="0" precision="FP32" />
 			</output>
 		</layer>
-		<layer id="5286" name="Constant_277063" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2790" name="Constant_52940" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5287" name="__module.model.gpt_neox.layers.21.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5288" name="__module.model.gpt_neox.layers.21.attention/aten::size/Constant_7" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2791" name="Constant_52942" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="5289" name="__module.model.gpt_neox.layers.21.attention/aten::size/Gather_7" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="3507,3512" />
-			</output>
-		</layer>
-		<layer id="5290" name="__module.model.gpt_neox.layers.21.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" />
-			</output>
-		</layer>
-		<layer id="5291" name="__module.model.gpt_neox.layers.21.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-			</input>
-			<output>
-				<port id="1" precision="I64" names="3509,3510,3514,3515" />
-			</output>
-		</layer>
-		<layer id="5292" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="5293" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5294" name="Constant_277062" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2792" name="Constant_52944" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5295" name="ScatterUpdate_277065" type="ScatterUpdate" version="opset3">
+		<layer id="2793" name="Concat_52945" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -80931,7 +42262,7 @@
 				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
@@ -80941,318 +42272,318 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5296" name="Constant_277068" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
+		<layer id="2794" name="Broadcast_52948" type="Broadcast" version="opset3">
+			<data mode="numpy" />
+			<input>
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="5297" name="Constant_277071" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5298" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2795" name="ReadValue_50217" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.16.valuepresent.16.value" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3516,x2.87">
+				<port id="1" precision="FP32" names="past_key_values.16.value">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5299" name="Constant_299287_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5300" name="Constant_299287" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
+		<layer id="2796" name="Constant_47877" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="5301" name="__module.model.gpt_neox.layers.21.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2797" name="Gather_47878" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="1" precision="I32">
+					<dim>-1</dim>
 				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3517">
+				<port id="3" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5302" name="Constant_277116" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2798" name="Constant_10832" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5303" name="Constant_277121" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2799" name="Constant_10834" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5304" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2800" name="Constant_10836" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2801" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Slice_10" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3511">
+				<port id="5" precision="FP32" names="2195">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5305" name="__module.model.gpt_neox.layers.21.attention/aten::cat/Concat_2" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="2802" name="Constant_10845" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
+			<output>
+				<port id="0" precision="I64" names="2196">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2803" name="__module.model.gpt_neox.layers.16.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3519">
+				<port id="2" precision="FP32" names="2197">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5306" name="__module.model.gpt_neox.layers.21.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2804" name="__module.model.gpt_neox.layers.16.attention/aten::cat/Concat_5" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3520">
+				<port id="2" precision="FP32" names="present.16.value">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5307" name="__module.model.gpt_neox.layers.21.attention/aten::add/Add_1" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2805" name="Constant_11227" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2806" name="ShapeOf_39848" type="ShapeOf" version="opset3">
+			<data output_type="i64" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3521">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5308" name="Constant_277166" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2807" name="Constant_39849" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116168" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5309" name="Constant_277169" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2808" name="Constant_39850" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
+				<port id="0" precision="I64" />
+			</output>
+		</layer>
+		<layer id="2809" name="Gather_39851" type="Gather" version="opset8">
+			<data batch_dims="0" />
+			<input>
 				<port id="0" precision="I64">
 					<dim>4</dim>
 				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64" />
+			</input>
+			<output>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5310" name="Constant_277172" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2810" name="Add_39854" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="I64" names="2244">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2811" name="Constant_11230" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5311" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2812" name="__module.model.gpt_neox.layers.16.attention/aten::slice/Slice_14" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3469">
+				<port id="5" precision="FP32" names="2248">
+					<dim>-1</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>60</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5312" name="__module.model.gpt_neox.layers.21.attention/aten::cat/Concat_3" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="2813" name="__module.model.gpt_neox.layers.16.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+			<data causal="false" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>60</dim>
+					<dim>80</dim>
+				</port>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="3" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3525">
+				<port id="4" precision="FP32" names="2249,attn_output.97">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -81260,8 +42591,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5313" name="__module.model.gpt_neox.layers.21.attention/aten::to/Convert_1" type="Concat" version="opset1">
-			<data axis="-2" />
+		<layer id="2814" name="__module.model.gpt_neox.layers.16.attention/aten::transpose/Constant" type="Const" version="opset1">
+			<data element_type="i32" shape="4" offset="201116200" size="16" />
+			<output>
+				<port id="0" precision="I32">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2815" name="__module.model.gpt_neox.layers.16.attention/aten::transpose/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -81269,869 +42607,1044 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+				<port id="1" precision="I32">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="present.21.key">
+				<port id="2" precision="FP32" names="2250">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5314" name="Constant_337988_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
+		<layer id="2816" name="Constant_41000" type="Const" version="opset1">
+			<data element_type="i64" shape="3" offset="201116216" size="24" />
 			<output>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="I64">
+					<dim>3</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5315" name="Constant_337988" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="2817" name="__module.model.gpt_neox.layers.16.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
+			<data special_zero="true" />
 			<input>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>3</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" />
+				<port id="2" precision="FP32" names="2253">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5316" name="Constant_337981" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
+		<layer id="2818" name="self.model.gpt_neox.layers.16.attention.dense.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 2560" offset="1462233680" size="6553600" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5317" name="Constant_337983" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
+		<layer id="2819" name="Convert_75352" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5318" name="Constant_337985" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
+		<layer id="2820" name="self.model.gpt_neox.layers.16.attention.dense.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="1468787280" size="2560" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="U8">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5319" name="Concat_337986" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="2821" name="Convert_75355" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64">
+				<port id="0" precision="U8">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64">
+			</input>
+			<output>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+			</output>
+		</layer>
+		<layer id="2822" name="self.model.gpt_neox.layers.16.attention.dense.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5320" name="Broadcast_337989" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
+		<layer id="2823" name="self.model.gpt_neox.layers.16.attention.dense.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="1468789840" size="5120" />
 			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5321" name="ReadValue_330483" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.21.valuepresent.21.value" variable_type="f32" variable_shape="?,32,?,80" />
+		<layer id="2824" name="self.model.gpt_neox.layers.16.attention.dense.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="past_key_values.21.value">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5322" name="Constant_323514" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
+		<layer id="2825" name="self.model.gpt_neox.layers.16.attention.dense.weight/fq_weights_1/convert" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64" />
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5323" name="Gather_323515" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="2826" name="__module.model.gpt_neox.layers.16.attention.dense/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="3" precision="FP32">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5324" name="Constant_277178" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5325" name="Constant_277181" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5326" name="Constant_277184" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2827" name="Constant_37423" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1468794960" size="10240" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5327" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2828" name="__module.model.gpt_neox.layers.16.attention.dense/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3463">
+				<port id="2" precision="FP32" names="2256,2258,input.65">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5328" name="Constant_12452" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="3464">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5329" name="__module.model.gpt_neox.layers.21.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
+		<layer id="2829" name="__module.model.gpt_neox.layers.16/aten::add/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3465">
+				<port id="2" precision="FP32" names="2276">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5330" name="__module.model.gpt_neox.layers.21.attention/aten::cat/Concat_5" type="Concat" version="opset1">
-			<data axis="-2" />
+		<layer id="2830" name="__module.model.gpt_neox.layers.16/aten::add/Add_1" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="present.21.value">
+				<port id="2" precision="FP32" names="2277,2279,hidden_states.71">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5331" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Slice_11" type="Const" version="opset1">
-			<data element_type="boolean" shape="1, 1, 2048, 2048" offset="201244276" size="4194304" />
+		<layer id="2831" name="__module.model.gpt_neox.layers.17.input_layernorm/aten::layer_norm/Multiply" type="Const" version="opset1">
+			<data element_type="i32" shape="1" offset="128867668" size="4" />
 			<output>
-				<port id="0" precision="BOOL" names="3542">
-					<dim>1</dim>
+				<port id="0" precision="I32">
 					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5332" name="Constant_277188" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2832" name="__module.model.gpt_neox.layers.17.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
+			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="I32">
+					<dim>1</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5333" name="Constant_277187" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162298" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2833" name="Constant_37424" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1468805200" size="10240" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="FP32">
+					<dim>1</dim>
 					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5334" name="__module.model.gpt_neox.layers.21.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="2834" name="__module.model.gpt_neox.layers.17.post_attention_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5335" name="__module.model.gpt_neox.layers.21.attention/aten::size/Constant_8" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2835" name="Constant_37425" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1468815440" size="10240" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5336" name="__module.model.gpt_neox.layers.21.attention/aten::size/Gather_8" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="2836" name="__module.model.gpt_neox.layers.17.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="3537" />
+				<port id="2" precision="FP32" names="2377">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5337" name="__module.model.gpt_neox.layers.21.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="2837" name="self.model.gpt_neox.layers.17.mlp.dense_h_to_4h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 2560" offset="1468825680" size="26214400" />
+			<output>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2838" name="Convert_74524" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5338" name="__module.model.gpt_neox.layers.21.attention/aten::size/Constant_9" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2839" name="self.model.gpt_neox.layers.17.mlp.dense_h_to_4h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 1" offset="1495040080" size="10240" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5339" name="__module.model.gpt_neox.layers.21.attention/aten::size/Gather_9" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="2840" name="Convert_74527" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="3535" />
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5340" name="__module.model.gpt_neox.layers.21.attention/aten::sub/Subtract" type="Subtract" version="opset1">
+		<layer id="2841" name="self.model.gpt_neox.layers.17.mlp.dense_h_to_4h.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64" names="3539,3540" />
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5341" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Constant_4" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2842" name="self.model.gpt_neox.layers.17.mlp.dense_h_to_4h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="10240, 1" offset="1495050320" size="20480" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5342" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1">
+		<layer id="2843" name="self.model.gpt_neox.layers.17.mlp.dense_h_to_4h.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5343" name="Constant_277186" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
+		<layer id="2844" name="__module.model.gpt_neox.layers.17.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<rt_info>
-				<attribute name="precise" version="0" />
+				<attribute name="decompression" version="0" />
 			</rt_info>
+			<input>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5344" name="ScatterUpdate_277189" type="ScatterUpdate" version="opset3">
+		<layer id="2845" name="__module.model.gpt_neox.layers.17.mlp.dense_h_to_4h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5345" name="Constant_277190" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2846" name="self.model.gpt_neox.layers.17.mlp.dense_h_to_4h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="10240" offset="1495070800" size="20480" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.17.mlp.dense_h_to_4h.bias">
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5346" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1">
+		<layer id="2847" name="__module.model.gpt_neox.layers.17.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<rt_info>
+				<attribute name="decompression" version="0" />
+			</rt_info>
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
+				<port id="0" precision="BF16">
+					<dim>10240</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5347" name="ScatterUpdate_277191" type="ScatterUpdate" version="opset3">
+		<layer id="2848" name="__module.model.gpt_neox.layers.17.mlp.dense_h_to_4h/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5348" name="Constant_277194" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438612" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="2" precision="FP32" names="2382">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5349" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2849" name="__module.model.gpt_neox.layers.17.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
+			<data approximation_mode="ERF" />
 			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>3</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="BOOL" names="3543">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="1" precision="FP32" names="2383">
 					<dim>-1</dim>
-					<dim>2048</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5350" name="Constant_277267" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2850" name="self.model.gpt_neox.layers.17.mlp.dense_4h_to_h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 10240" offset="1495091280" size="26214400" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5351" name="Constant_277268" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2851" name="Convert_74534" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5352" name="Constant_277264" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2852" name="self.model.gpt_neox.layers.17.mlp.dense_4h_to_h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="1521305680" size="2560" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="U8">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5353" name="Constant_277263" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2853" name="Convert_74537" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I32">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5354" name="ScatterUpdate_277269" type="ScatterUpdate" version="opset3">
+		<layer id="2854" name="self.model.gpt_neox.layers.17.mlp.dense_4h_to_h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2855" name="self.model.gpt_neox.layers.17.mlp.dense_4h_to_h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="1521308240" size="5120" />
+			<output>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+			</output>
+		</layer>
+		<layer id="2856" name="self.model.gpt_neox.layers.17.mlp.dense_4h_to_h.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5355" name="Constant_277272" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
+		<layer id="2857" name="__module.model.gpt_neox.layers.17.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<rt_info>
-				<attribute name="precise" version="0" />
+				<attribute name="decompression" version="0" />
 			</rt_info>
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5356" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2858" name="__module.model.gpt_neox.layers.17.mlp.dense_4h_to_h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="BOOL" names="3544,3545">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5357" name="__module.model.gpt_neox.layers.21.attention/aten::where/Convert_1_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
+		<layer id="2859" name="self.model.gpt_neox.layers.17.mlp.dense_4h_to_h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="2560" offset="1521313360" size="5120" />
 			<output>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.17.mlp.dense_4h_to_h.bias">
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5358" name="__module.model.gpt_neox.layers.21.attention/aten::where/Convert_1" type="Convert" version="opset1">
+		<layer id="2860" name="__module.model.gpt_neox.layers.17.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
 			</rt_info>
 			<input>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="BF16">
+					<dim>2560</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="5359" name="__module.model.gpt_neox.layers.21.attention/aten::full/Convert" type="Const" version="opset1">
-			<data element_type="f32" shape="" offset="205438636" size="4" />
-			<output>
-				<port id="0" precision="FP32" names="3532" />
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5360" name="__module.model.gpt_neox.layers.21.attention/aten::where/Select" type="Select" version="opset1">
+		<layer id="2861" name="__module.model.gpt_neox.layers.17.mlp.dense_4h_to_h/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="FP32" />
-				<port id="2" precision="FP32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="3546,causal_mask.85">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="2386,input.71">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5361" name="Add_290464" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
+		<layer id="2862" name="Constant_37426" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1521318480" size="10240" />
+			<output>
 				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2863" name="__module.model.gpt_neox.layers.17.input_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5362" name="ShapeOf_290465" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="2864" name="Constant_37427" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1521328720" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2865" name="__module.model.gpt_neox.layers.17.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
 					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="2289,hidden_states.69">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2866" name="self.model.gpt_neox.layers.17.attention.query_key_value.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 2560" offset="1521338960" size="19660800" />
+			<output>
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2867" name="Convert_75011" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5363" name="Constant_254644" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2868" name="self.model.gpt_neox.layers.17.attention.query_key_value.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 1" offset="1540999760" size="7680" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="U8">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5364" name="__module.model.gpt_neox.layers.21.attention/aten::size/Constant_10" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2869" name="Convert_75014" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>1</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I32" />
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5365" name="__module.model.gpt_neox.layers.21.attention/aten::size/Gather_10" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="2870" name="self.model.gpt_neox.layers.17.attention.query_key_value.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="3530">
+				<port id="2" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2871" name="self.model.gpt_neox.layers.17.attention.query_key_value.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="7680, 1" offset="1541007440" size="15360" />
+			<output>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5366" name="__module.model.gpt_neox.layers.21.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="2872" name="self.model.gpt_neox.layers.17.attention.query_key_value.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
+			</output>
+		</layer>
+		<layer id="2873" name="self.model.gpt_neox.layers.17.attention.query_key_value.weight/fq_weights_1/convert" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64" names="3547">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5367" name="__module.model.gpt_neox.layers.21.attention/aten::expand/Abs" type="Abs" version="opset1">
+		<layer id="2874" name="__module.model.gpt_neox.layers.17.attention.query_key_value/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>7680</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2875" name="Constant_37428" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 7680" offset="1541022800" size="30720" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5368" name="Maximum_290466" type="Maximum" version="opset1">
+		<layer id="2876" name="__module.model.gpt_neox.layers.17.attention.query_key_value/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>7680</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>7680</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
+				<port id="2" precision="FP32" names="2296,qkv.69">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>7680</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2877" name="__module.model.gpt_neox.layers.17.attention/prim::ListConstruct" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201115992" size="32" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5369" name="__module.model.gpt_neox.layers.21.attention/aten::add/Add_2" type="Broadcast" version="opset3">
-			<data mode="numpy" />
+		<layer id="2878" name="__module.model.gpt_neox.layers.17.attention/aten::view/Reshape" type="Reshape" version="opset1">
+			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>7680</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3549">
-					<dim>-1</dim>
-					<dim>-1</dim>
+				<port id="2" precision="FP32" names="2300,qkv.71">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5370" name="__module.model.gpt_neox.layers.21.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
-			<data causal="false" />
+		<layer id="2879" name="Constant_11426" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2880" name="Constant_11428" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2881" name="Constant_11430" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2882" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Slice" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>240</dim>
 				</port>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+				<port id="1" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3550,3551,sdpa_result.43,tensor.85">
+				<port id="5" precision="FP32" names="2301">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5371" name="Constant_12893" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="2883" name="Constant_11439" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64" names="3552">
+				<port id="0" precision="I64" names="2302">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5372" name="__module.model.gpt_neox.layers.21.attention/aten::permute/Transpose_3" type="Transpose" version="opset1">
+		<layer id="2884" name="__module.model.gpt_neox.layers.17.attention/aten::permute/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
@@ -82139,1154 +43652,1135 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3553">
-					<dim>-1</dim>
+				<port id="2" precision="FP32" names="2303,query.35">
 					<dim>-1</dim>
 					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5373" name="__module.model.gpt_neox.layers.21.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438650" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2885" name="Constant_11480" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>3</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5374" name="__module.model.gpt_neox.layers.21.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
-			<data special_zero="true" />
+		<layer id="2886" name="Constant_11482" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2887" name="Constant_11484" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2888" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Slice_1" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>3</dim>
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3558">
+				<port id="5" precision="FP32" names="2310,q.35">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5375" name="self.model.gpt_neox.layers.21.attention.dense.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 2560" offset="1859795154" size="6553600" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5376" name="Convert_22274" type="Convert" version="opset1">
-			<data destination_type="f16" />
+		<layer id="2889" name="__module.model.gpt_neox.layers.17.attention/aten::mul/Multiply" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32" names="2316">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5377" name="self.model.gpt_neox.layers.21.attention.dense.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="1866348754" size="2560" />
+		<layer id="2890" name="Constant_11638" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5378" name="Convert_22277" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
+		<layer id="2891" name="Constant_11640" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="2892" name="Constant_11642" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5379" name="self.model.gpt_neox.layers.21.attention.dense.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2893" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Slice_2" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5380" name="self.model.gpt_neox.layers.21.attention.dense.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="1866351314" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
+				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="5381" name="self.model.gpt_neox.layers.21.attention.dense.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="3" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="5" precision="FP32" names="2326,x2.69">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5382" name="self.model.gpt_neox.layers.21.attention.dense.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="2894" name="Constant_37429" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5383" name="__module.model.gpt_neox.layers.21.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
+		<layer id="2895" name="__module.model.gpt_neox.layers.17.attention/aten::neg/Multiply" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>2560</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="2327">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5384" name="Constant_299288_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1866356434" size="5120" />
+		<layer id="2896" name="Constant_11535" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5385" name="Constant_299288" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+		<layer id="2897" name="Constant_11579" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="2898" name="Constant_11581" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5386" name="__module.model.gpt_neox.layers.21.attention.dense/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2899" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Slice_3" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="FP32">
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3561,3563,input.87">
+				<port id="5" precision="FP32" names="2321">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5387" name="__module.model.gpt_neox.layers.21/aten::add/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2900" name="__module.model.gpt_neox.layers.17.attention/aten::cat/Concat" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3581">
+				<port id="2" precision="FP32" names="2329">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5388" name="__module.model.gpt_neox.layers.21/aten::add/Add_1" type="Add" version="opset1">
+		<layer id="2901" name="__module.model.gpt_neox.layers.17.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3582,3584,hidden_states.45">
+				<port id="2" precision="FP32" names="2330">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5389" name="__module.model.gpt_neox.layers.22.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867670" size="4" />
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5390" name="__module.model.gpt_neox.layers.22.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
-			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
+		<layer id="2902" name="__module.model.gpt_neox.layers.17.attention/aten::add/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="I32">
-					<dim>1</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="2331">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5391" name="Constant_299289_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1866361554" size="5120" />
+		<layer id="2903" name="Constant_11490" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5392" name="Constant_299289" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+		<layer id="2904" name="Constant_11492" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="2905" name="Constant_11494" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5393" name="__module.model.gpt_neox.layers.22.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2906" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Slice_4" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="1" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="5394" name="Constant_299290_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1866366674" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="2" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="5395" name="Constant_299290" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
+				<port id="3" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+				<port id="5" precision="FP32" names="2311">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5396" name="__module.model.gpt_neox.layers.22.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2907" name="__module.model.gpt_neox.layers.17.attention/aten::cat/Concat_1" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3719">
+				<port id="2" precision="FP32" names="2349">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5397" name="self.model.gpt_neox.layers.22.mlp.dense_h_to_4h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 2560" offset="1866371794" size="26214400" />
+		<layer id="2908" name="Constant_52959" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
+				<port id="0" precision="FP32" />
 			</output>
 		</layer>
-		<layer id="5398" name="Convert_22324" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="2909" name="Constant_52952" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5399" name="self.model.gpt_neox.layers.22.mlp.dense_h_to_4h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 1" offset="1892586194" size="10240" />
+		<layer id="2910" name="Constant_52954" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5400" name="Convert_22327" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</input>
+		<layer id="2911" name="Constant_52956" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5401" name="self.model.gpt_neox.layers.22.mlp.dense_h_to_4h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2912" name="Concat_52957" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
+				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="2" precision="I64">
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="5402" name="self.model.gpt_neox.layers.22.mlp.dense_h_to_4h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="10240, 1" offset="1892596434" size="20480" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
+			</input>
+			<output>
+				<port id="4" precision="I64">
+					<dim>4</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5403" name="self.model.gpt_neox.layers.22.mlp.dense_h_to_4h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2913" name="Broadcast_52960" type="Broadcast" version="opset3">
+			<data mode="numpy" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5404" name="self.model.gpt_neox.layers.22.mlp.dense_h_to_4h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="2914" name="ReadValue_50219" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.17.keypresent.17.key" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="1" precision="FP32" names="past_key_values.17.key">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5405" name="__module.model.gpt_neox.layers.22.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
+		<layer id="2915" name="Constant_47880" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64" />
+			</output>
+		</layer>
+		<layer id="2916" name="Gather_47881" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="1" precision="I32">
+					<dim>-1</dim>
 				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="3" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5406" name="Constant_299291_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 10240" offset="1892616914" size="20480" />
+		<layer id="2917" name="Constant_11444" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5407" name="Constant_299291" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+		<layer id="2918" name="Constant_11446" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>10240</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="2919" name="Constant_11448" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5408" name="__module.model.gpt_neox.layers.22.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2920" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Slice_5" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
-				<port id="1" precision="FP32">
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3724">
+				<port id="5" precision="FP32" names="2304">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5409" name="__module.model.gpt_neox.layers.22.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
-			<data approximation_mode="ERF" />
+		<layer id="2921" name="Constant_11457" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
+			<output>
+				<port id="0" precision="I64" names="2305">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2922" name="__module.model.gpt_neox.layers.17.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="3725">
+				<port id="2" precision="FP32" names="2306,key.69">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5410" name="self.model.gpt_neox.layers.22.mlp.dense_4h_to_h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 10240" offset="1892637394" size="26214400" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5411" name="Convert_22334" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
+		<layer id="2923" name="Constant_11500" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5412" name="self.model.gpt_neox.layers.22.mlp.dense_4h_to_h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="1918851794" size="2560" />
+		<layer id="2924" name="Constant_11502" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5413" name="Convert_22337" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
+		<layer id="2925" name="Constant_11504" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5414" name="self.model.gpt_neox.layers.22.mlp.dense_4h_to_h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2926" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Slice_6" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5415" name="self.model.gpt_neox.layers.22.mlp.dense_4h_to_h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="1918854354" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
+				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="5416" name="self.model.gpt_neox.layers.22.mlp.dense_4h_to_h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5417" name="self.model.gpt_neox.layers.22.mlp.dense_4h_to_h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
+			<output>
+				<port id="5" precision="FP32" names="2312,k.35">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5418" name="__module.model.gpt_neox.layers.22.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
+		<layer id="2927" name="__module.model.gpt_neox.layers.17.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="2332">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5419" name="Constant_299292_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1918859474" size="5120" />
+		<layer id="2928" name="Constant_11781" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5420" name="Constant_299292" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+		<layer id="2929" name="Constant_11783" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="2930" name="Constant_11785" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5421" name="__module.model.gpt_neox.layers.22.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2931" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Slice_7" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="FP32">
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3728,input.93">
+				<port id="5" precision="FP32" names="2342,x2.71">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5422" name="Constant_299293_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1918864594" size="5120" />
+		<layer id="2932" name="Constant_37430" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5423" name="Constant_299293" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5424" name="__module.model.gpt_neox.layers.22.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
+		<layer id="2933" name="__module.model.gpt_neox.layers.17.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="2343">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5425" name="Constant_299294_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1918869714" size="5120" />
+		<layer id="2934" name="Constant_11678" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="FP16">
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2935" name="Constant_11722" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2936" name="Constant_11724" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5426" name="Constant_299294" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="2937" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Slice_8" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
+				<port id="3" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="5" precision="FP32" names="2337">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5427" name="__module.model.gpt_neox.layers.22.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2938" name="__module.model.gpt_neox.layers.17.attention/aten::cat/Concat_2" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3594">
+				<port id="2" precision="FP32" names="2345">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5428" name="self.model.gpt_neox.layers.22.attention.query_key_value.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 2560" offset="1918874834" size="19660800" />
+		<layer id="2939" name="__module.model.gpt_neox.layers.17.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32" names="2346">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5429" name="Convert_22304" type="Convert" version="opset1">
-			<data destination_type="f16" />
+		<layer id="2940" name="__module.model.gpt_neox.layers.17.attention/aten::add/Add_1" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32" names="2347">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5430" name="self.model.gpt_neox.layers.22.attention.query_key_value.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 1" offset="1938535634" size="7680" />
+		<layer id="2941" name="Constant_11510" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5431" name="Convert_22307" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
+		<layer id="2942" name="Constant_11512" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="2943" name="Constant_11514" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5432" name="self.model.gpt_neox.layers.22.attention.query_key_value.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2944" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Slice_9" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
+				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5433" name="self.model.gpt_neox.layers.22.attention.query_key_value.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="7680, 1" offset="1938543314" size="15360" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
+				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="5434" name="self.model.gpt_neox.layers.22.attention.query_key_value.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="3" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="5" precision="FP32" names="2313">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5435" name="self.model.gpt_neox.layers.22.attention.query_key_value.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="2945" name="__module.model.gpt_neox.layers.17.attention/aten::cat/Concat_3" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32" names="2351">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5436" name="__module.model.gpt_neox.layers.22.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
+		<layer id="2946" name="__module.model.gpt_neox.layers.17.attention/aten::cat/Concat_4" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>7680</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="present.17.key">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5437" name="Constant_299295_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 7680" offset="1938558674" size="15360" />
+		<layer id="2947" name="Constant_52971" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32" />
+			</output>
+		</layer>
+		<layer id="2948" name="Constant_52964" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2949" name="Constant_52966" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2950" name="Constant_52968" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5438" name="Constant_299295" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="2951" name="Concat_52969" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
 					<dim>1</dim>
-					<dim>7680</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
+				<port id="4" precision="I64">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5439" name="__module.model.gpt_neox.layers.22.attention.query_key_value/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="2952" name="Broadcast_52972" type="Broadcast" version="opset3">
+			<data mode="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
+					<dim>4</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
+			</output>
+		</layer>
+		<layer id="2953" name="ReadValue_50221" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.17.valuepresent.17.value" variable_type="f32" variable_shape="?,32,?,80" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3601,qkv.89">
+				<port id="1" precision="FP32" names="past_key_values.17.value">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5440" name="__module.model.gpt_neox.layers.22.attention/prim::ListConstruct/Concat" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080154" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2954" name="Constant_47883" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="5441" name="__module.model.gpt_neox.layers.22.attention/aten::view/Reshape" type="Reshape" version="opset1">
-			<data special_zero="true" />
+		<layer id="2955" name="Gather_47884" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="I32">
+					<dim>-1</dim>
 				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3605,qkv.91">
-					<dim>-1</dim>
+				<port id="3" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>240</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5442" name="Constant_277279" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2956" name="Constant_11462" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5443" name="Constant_277282" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2957" name="Constant_11464" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5444" name="Constant_277285" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2958" name="Constant_11466" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5445" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2959" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Slice_10" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -83295,17 +44789,20 @@
 					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3606">
+				<port id="5" precision="FP32" names="2307">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -83313,15 +44810,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5446" name="Constant_13005" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="2960" name="Constant_11475" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64" names="3607">
+				<port id="0" precision="I64" names="2308">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5447" name="__module.model.gpt_neox.layers.22.attention/aten::permute/Transpose" type="Transpose" version="opset1">
+		<layer id="2961" name="__module.model.gpt_neox.layers.17.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -83334,7 +44831,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3608,query.133">
+				<port id="2" precision="FP32" names="2309">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -83342,41 +44839,41 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5448" name="Constant_277291" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+		<layer id="2962" name="__module.model.gpt_neox.layers.17.attention/aten::cat/Concat_5" type="Concat" version="opset1">
+			<data axis="-2" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="5449" name="Constant_277294" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="present.17.value">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5450" name="Constant_277297" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2963" name="Constant_11857" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5451" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2964" name="ShapeOf_39919" type="ShapeOf" version="opset3">
+			<data output_type="i64" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -83384,1053 +44881,1197 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
+			</input>
+			<output>
 				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="2965" name="Constant_39920" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116168" size="8" />
 			<output>
-				<port id="4" precision="FP32" names="3615,q.45">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5452" name="self.model.gpt_neox.layers.22.attention.rotary_emb.cos_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201080346" size="81920" />
+		<layer id="2966" name="Constant_39921" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.22.attention.rotary_emb.cos_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="5453" name="self.model.gpt_neox.layers.22.attention.rotary_emb.cos_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="2967" name="Gather_39922" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="I64">
+					<dim>4</dim>
 				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="3" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5454" name="__module.model.gpt_neox.layers.22.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
+		<layer id="2968" name="Add_39925" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="I64" names="2356">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5455" name="Constant_277303" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2969" name="Constant_11860" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5456" name="Constant_277306" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+		<layer id="2970" name="__module.model.gpt_neox.layers.17.attention/aten::slice/Slice_14" type="Slice" version="opset8">
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="5457" name="Constant_277309" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="5" precision="FP32" names="2360">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5458" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2971" name="__module.model.gpt_neox.layers.17.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+			<data causal="false" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>240</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="3" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3609">
-					<dim>-1</dim>
+				<port id="4" precision="FP32" names="2361,attn_output.103">
 					<dim>-1</dim>
 					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5459" name="Constant_13019" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="2972" name="__module.model.gpt_neox.layers.17.attention/aten::transpose/Constant" type="Const" version="opset1">
+			<data element_type="i32" shape="4" offset="201116200" size="16" />
 			<output>
-				<port id="0" precision="I64" names="3610">
+				<port id="0" precision="I32">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5460" name="__module.model.gpt_neox.layers.22.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
+		<layer id="2973" name="__module.model.gpt_neox.layers.17.attention/aten::transpose/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="I32">
 					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3611,key.133">
+				<port id="2" precision="FP32" names="2362">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5461" name="__module.model.gpt_neox.layers.22.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="2974" name="Constant_41001" type="Const" version="opset1">
+			<data element_type="i64" shape="3" offset="201116216" size="24" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>3</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2975" name="__module.model.gpt_neox.layers.17.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
+			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>80</dim>
 				</port>
+				<port id="1" precision="I64">
+					<dim>3</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="2365">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5462" name="__module.model.gpt_neox.layers.22.attention/aten::size/Constant_2" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2976" name="self.model.gpt_neox.layers.17.attention.dense.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 2560" offset="1541053520" size="6553600" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5463" name="__module.model.gpt_neox.layers.22.attention/aten::size/Gather_2" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="2977" name="Convert_75363" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="3619" />
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5464" name="Constant_338001_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
+		<layer id="2978" name="self.model.gpt_neox.layers.17.attention.dense.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="1547607120" size="2560" />
 			<output>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5465" name="Constant_338001" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="2979" name="Convert_75366" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="5466" name="Constant_337994" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
-			<output>
-				<port id="0" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5467" name="Constant_337996" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64">
+		<layer id="2980" name="self.model.gpt_neox.layers.17.attention.dense.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5468" name="Constant_337998" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
+		<layer id="2981" name="self.model.gpt_neox.layers.17.attention.dense.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="1547609680" size="5120" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5469" name="Concat_337999" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="2982" name="self.model.gpt_neox.layers.17.attention.dense.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
+			</output>
+		</layer>
+		<layer id="2983" name="self.model.gpt_neox.layers.17.attention.dense.weight/fq_weights_1/convert" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5470" name="Broadcast_338002" type="Broadcast" version="opset3">
-			<data mode="numpy" />
+		<layer id="2984" name="__module.model.gpt_neox.layers.17.attention.dense/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5471" name="ReadValue_330485" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.22.keypresent.22.key" variable_type="f32" variable_shape="?,32,?,80" />
+		<layer id="2985" name="Constant_37431" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1547614800" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2986" name="__module.model.gpt_neox.layers.17.attention.dense/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="past_key_values.22.key">
+				<port id="2" precision="FP32" names="2368,2370,input.69">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5472" name="Constant_323517" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
+		<layer id="2987" name="__module.model.gpt_neox.layers.17/aten::add/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64" />
+				<port id="2" precision="FP32" names="2388">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5473" name="Gather_323518" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="2988" name="__module.model.gpt_neox.layers.17/aten::add/Add_1" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I32">
+				<port id="1" precision="FP32">
 					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="3" precision="FP32">
+				<port id="2" precision="FP32" names="2389,2391,hidden_states.75">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5474" name="__module.model.gpt_neox.layers.22.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="2989" name="__module.model.gpt_neox.layers.18.input_layernorm/aten::layer_norm/Multiply" type="Const" version="opset1">
+			<data element_type="i32" shape="1" offset="128867668" size="4" />
+			<output>
+				<port id="0" precision="I32">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2990" name="__module.model.gpt_neox.layers.18.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
+			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="I32">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5475" name="__module.model.gpt_neox.layers.22.attention/aten::size/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2991" name="Constant_37432" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1547625040" size="10240" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5476" name="__module.model.gpt_neox.layers.22.attention/aten::size/Gather_3" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="2992" name="__module.model.gpt_neox.layers.18.post_attention_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="3621" />
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="2993" name="Constant_37433" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1547635280" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5477" name="__module.model.gpt_neox.layers.22.attention/aten::add_/Add" type="Add" version="opset1">
+		<layer id="2994" name="__module.model.gpt_neox.layers.18.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64" names="3620,3624,3625,seq_len.89" />
+				<port id="2" precision="FP32" names="2489">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5478" name="__module.model.gpt_neox.layers.22.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2995" name="self.model.gpt_neox.layers.18.mlp.dense_h_to_4h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 2560" offset="1547645520" size="26214400" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5479" name="__module.model.gpt_neox.layers.22.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1">
+		<layer id="2996" name="Convert_74544" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5480" name="__module.model.gpt_neox.layers.22.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="2997" name="self.model.gpt_neox.layers.18.mlp.dense_h_to_4h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 1" offset="1573859920" size="10240" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="U8">
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5481" name="__module.model.gpt_neox.layers.22.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="2998" name="Convert_74547" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
+				<port id="0" precision="U8">
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
+			</input>
+			<output>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I64">
+			</output>
+		</layer>
+		<layer id="2999" name="self.model.gpt_neox.layers.18.mlp.dense_h_to_4h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3628,3631,cos.89">
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5482" name="Constant_38329" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
+		<layer id="3000" name="self.model.gpt_neox.layers.18.mlp.dense_h_to_4h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="10240, 1" offset="1573870160" size="20480" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5483" name="__module.model.gpt_neox.layers.22.attention/aten::index/Gather" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="3001" name="self.model.gpt_neox.layers.18.mlp.dense_h_to_4h.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
 				</port>
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="3634">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5484" name="__module.model.gpt_neox.layers.22.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1">
+		<layer id="3002" name="__module.model.gpt_neox.layers.18.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<rt_info>
+				<attribute name="decompression" version="0" />
+			</rt_info>
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3635,cos.91">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5485" name="__module.model.gpt_neox.layers.22.attention/aten::mul/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3003" name="__module.model.gpt_neox.layers.18.mlp.dense_h_to_4h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3639">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5486" name="Constant_277314" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3004" name="self.model.gpt_neox.layers.18.mlp.dense_h_to_4h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="10240" offset="1573890640" size="20480" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.18.mlp.dense_h_to_4h.bias">
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5487" name="Constant_277313" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
+		<layer id="3005" name="__module.model.gpt_neox.layers.18.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<rt_info>
-				<attribute name="precise" version="0" />
+				<attribute name="decompression" version="0" />
 			</rt_info>
+			<input>
+				<port id="0" precision="BF16">
+					<dim>10240</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5488" name="__module.model.gpt_neox.layers.22.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="3006" name="__module.model.gpt_neox.layers.18.mlp.dense_h_to_4h/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>10240</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="2494">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5489" name="__module.model.gpt_neox.layers.22.attention/aten::size/Constant_5" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="5490" name="__module.model.gpt_neox.layers.22.attention/aten::size/Gather_5" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="3007" name="__module.model.gpt_neox.layers.18.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
+			<data approximation_mode="ERF" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="3640,3645" />
+				<port id="1" precision="FP32" names="2495">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5491" name="__module.model.gpt_neox.layers.22.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
+		<layer id="3008" name="self.model.gpt_neox.layers.18.mlp.dense_4h_to_h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 10240" offset="1573911120" size="26214400" />
 			<output>
-				<port id="2" precision="I64" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5492" name="__module.model.gpt_neox.layers.22.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1">
+		<layer id="3009" name="Convert_74554" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64" names="3642,3643,3647,3648" />
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5493" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Constant_1" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3010" name="self.model.gpt_neox.layers.18.mlp.dense_4h_to_h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="1600125520" size="2560" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5494" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1">
+		<layer id="3011" name="Convert_74557" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3012" name="self.model.gpt_neox.layers.18.mlp.dense_4h_to_h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5495" name="Constant_277312" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3013" name="self.model.gpt_neox.layers.18.mlp.dense_4h_to_h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="1600128080" size="5120" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5496" name="ScatterUpdate_277315" type="ScatterUpdate" version="opset3">
+		<layer id="3014" name="self.model.gpt_neox.layers.18.mlp.dense_4h_to_h.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5497" name="Constant_277318" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
+		<layer id="3015" name="__module.model.gpt_neox.layers.18.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<rt_info>
-				<attribute name="precise" version="0" />
+				<attribute name="decompression" version="0" />
 			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="5498" name="Constant_277321" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5499" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3016" name="__module.model.gpt_neox.layers.18.mlp.dense_4h_to_h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3649,x2.89">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5500" name="Constant_299296_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
+		<layer id="3017" name="self.model.gpt_neox.layers.18.mlp.dense_4h_to_h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="2560" offset="1600133200" size="5120" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.18.mlp.dense_4h_to_h.bias">
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5501" name="Constant_299296" type="Convert" version="opset1">
+		<layer id="3018" name="__module.model.gpt_neox.layers.18.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
 			</rt_info>
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16">
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5502" name="__module.model.gpt_neox.layers.22.attention/aten::neg/Multiply" type="Multiply" version="opset1">
+		<layer id="3019" name="__module.model.gpt_neox.layers.18.mlp.dense_4h_to_h/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3650">
+				<port id="2" precision="FP32" names="2498,input.75">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5503" name="Constant_277366" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5504" name="Constant_277371" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3020" name="Constant_37434" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1600138320" size="10240" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5505" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3021" name="__module.model.gpt_neox.layers.18.input_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3644">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5506" name="__module.model.gpt_neox.layers.22.attention/aten::cat/Concat" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="3022" name="Constant_37435" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1600148560" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3023" name="__module.model.gpt_neox.layers.18.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3652">
+				<port id="2" precision="FP32" names="2401,hidden_states.73">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5507" name="self.model.gpt_neox.layers.22.attention.rotary_emb.sin_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201162356" size="81920" />
+		<layer id="3024" name="self.model.gpt_neox.layers.18.attention.query_key_value.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 2560" offset="1600158800" size="19660800" />
 			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.22.attention.rotary_emb.sin_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5508" name="self.model.gpt_neox.layers.22.attention.rotary_emb.sin_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="3025" name="Convert_75022" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5509" name="__module.model.gpt_neox.layers.22.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3026" name="self.model.gpt_neox.layers.18.attention.query_key_value.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 1" offset="1619819600" size="7680" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="U8">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5510" name="__module.model.gpt_neox.layers.22.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3027" name="Convert_75025" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>1</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5511" name="__module.model.gpt_neox.layers.22.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3028" name="self.model.gpt_neox.layers.18.attention.query_key_value.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3029" name="self.model.gpt_neox.layers.18.attention.query_key_value.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="7680, 1" offset="1619827280" size="15360" />
+			<output>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I64">
+			</output>
+		</layer>
+		<layer id="3030" name="self.model.gpt_neox.layers.18.attention.query_key_value.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3629,3632,sin.89">
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="2" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5512" name="Constant_38332" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
+		<layer id="3031" name="self.model.gpt_neox.layers.18.attention.query_key_value.weight/fq_weights_1/convert" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I32" />
+				<port id="1" precision="FP32">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5513" name="__module.model.gpt_neox.layers.22.attention/aten::index/Gather_1" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="3032" name="__module.model.gpt_neox.layers.18.attention.query_key_value/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="3637">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5514" name="__module.model.gpt_neox.layers.22.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64" />
-			</input>
+		<layer id="3033" name="Constant_37436" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 7680" offset="1619842640" size="30720" />
 			<output>
-				<port id="2" precision="FP32" names="3638,sin.91">
-					<dim>-1</dim>
+				<port id="0" precision="FP32">
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>1</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5515" name="__module.model.gpt_neox.layers.22.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
+		<layer id="3034" name="__module.model.gpt_neox.layers.18.attention.query_key_value/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>7680</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>1</dim>
+					<dim>7680</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3653">
+				<port id="2" precision="FP32" names="2408,qkv.73">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5516" name="__module.model.gpt_neox.layers.22.attention/aten::add/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3035" name="__module.model.gpt_neox.layers.18.attention/prim::ListConstruct" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201115992" size="32" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3036" name="__module.model.gpt_neox.layers.18.attention/aten::view/Reshape" type="Reshape" version="opset1">
+			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>7680</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3654">
+				<port id="2" precision="FP32" names="2412,qkv.75">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5517" name="Constant_277417" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3037" name="Constant_12056" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5518" name="Constant_277420" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3038" name="Constant_12058" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5519" name="Constant_277423" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3039" name="Constant_12060" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5520" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3040" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Slice" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3616">
+				<port id="5" precision="FP32" names="2413">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>60</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5521" name="__module.model.gpt_neox.layers.22.attention/aten::cat/Concat_1" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="3041" name="Constant_12069" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
+			<output>
+				<port id="0" precision="I64" names="2414">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3042" name="__module.model.gpt_neox.layers.18.attention/aten::permute/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3672,3682,query.135,query.137">
+				<port id="2" precision="FP32" names="2415,query.37">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -84438,41 +46079,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5522" name="Constant_277429" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3043" name="Constant_12110" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5523" name="Constant_277432" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3044" name="Constant_12112" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5524" name="Constant_277435" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3045" name="Constant_12114" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5525" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3046" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Slice_1" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -84481,17 +46112,20 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3617,k.45">
+				<port id="5" precision="FP32" names="2422,q.37">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -84499,7 +46133,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5526" name="__module.model.gpt_neox.layers.22.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
+		<layer id="3047" name="__module.model.gpt_neox.layers.18.attention/aten::mul/Multiply" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -84516,7 +46150,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3655">
+				<port id="2" precision="FP32" names="2428">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -84524,160 +46158,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5527" name="Constant_277439" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5528" name="Constant_277438" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3048" name="Constant_12268" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5529" name="__module.model.gpt_neox.layers.22.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5530" name="__module.model.gpt_neox.layers.22.attention/aten::size/Constant_7" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3049" name="Constant_12270" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="5531" name="__module.model.gpt_neox.layers.22.attention/aten::size/Gather_7" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="3656,3661" />
-			</output>
-		</layer>
-		<layer id="5532" name="__module.model.gpt_neox.layers.22.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" />
-			</output>
-		</layer>
-		<layer id="5533" name="__module.model.gpt_neox.layers.22.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-			</input>
-			<output>
-				<port id="1" precision="I64" names="3658,3659,3663,3664" />
-			</output>
-		</layer>
-		<layer id="5534" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="5535" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5536" name="Constant_277437" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3050" name="Constant_12272" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5537" name="ScatterUpdate_277440" type="ScatterUpdate" version="opset3">
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5538" name="Constant_277443" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5539" name="Constant_277446" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
 			</output>
 		</layer>
-		<layer id="5540" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3051" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Slice_2" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -84686,17 +46191,20 @@
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3665,x2.91">
+				<port id="5" precision="FP32" names="2438,x2.73">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -84704,32 +46212,10 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5541" name="Constant_299297_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5542" name="Constant_299297" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
+		<layer id="3052" name="Constant_37437" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>1</dim>
@@ -84737,7 +46223,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5543" name="__module.model.gpt_neox.layers.22.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
+		<layer id="3053" name="__module.model.gpt_neox.layers.18.attention/aten::neg/Multiply" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -84754,7 +46240,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3666">
+				<port id="2" precision="FP32" names="2439">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -84762,30 +46248,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5544" name="Constant_277491" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3054" name="Constant_12165" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5545" name="Constant_277496" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3055" name="Constant_12209" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5546" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3056" name="Constant_12211" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3057" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Slice_3" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -84794,17 +46281,20 @@
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3660">
+				<port id="5" precision="FP32" names="2433">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -84812,7 +46302,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5547" name="__module.model.gpt_neox.layers.22.attention/aten::cat/Concat_2" type="Concat" version="opset1">
+		<layer id="3058" name="__module.model.gpt_neox.layers.18.attention/aten::cat/Concat" type="Concat" version="opset1">
 			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
@@ -84829,7 +46319,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3668">
+				<port id="2" precision="FP32" names="2441">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -84837,7 +46327,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5548" name="__module.model.gpt_neox.layers.22.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
+		<layer id="3059" name="__module.model.gpt_neox.layers.18.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -84854,7 +46344,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3669">
+				<port id="2" precision="FP32" names="2442">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -84862,7 +46352,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5549" name="__module.model.gpt_neox.layers.22.attention/aten::add/Add_1" type="Add" version="opset1">
+		<layer id="3060" name="__module.model.gpt_neox.layers.18.attention/aten::add/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -84879,7 +46369,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3670">
+				<port id="2" precision="FP32" names="2443">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -84887,41 +46377,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5550" name="Constant_277541" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3061" name="Constant_12120" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5551" name="Constant_277544" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3062" name="Constant_12122" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5552" name="Constant_277547" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3063" name="Constant_12124" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5553" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3064" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Slice_4" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -84930,17 +46410,20 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3618">
+				<port id="5" precision="FP32" names="2423">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -84948,7 +46431,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5554" name="__module.model.gpt_neox.layers.22.attention/aten::cat/Concat_3" type="Concat" version="opset1">
+		<layer id="3065" name="__module.model.gpt_neox.layers.18.attention/aten::cat/Concat_1" type="Concat" version="opset1">
 			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
@@ -84965,32 +46448,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3674">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5555" name="__module.model.gpt_neox.layers.22.attention/aten::to/Convert_1" type="Concat" version="opset1">
-			<data axis="-2" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="present.22.key">
+				<port id="2" precision="FP32" names="2461">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -84998,49 +46456,37 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5556" name="Constant_338014_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="5557" name="Constant_338014" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
+		<layer id="3066" name="Constant_52983" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="1" precision="FP32" />
+				<port id="0" precision="FP32" />
 			</output>
 		</layer>
-		<layer id="5558" name="Constant_338007" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
+		<layer id="3067" name="Constant_52976" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5559" name="Constant_338009" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
+		<layer id="3068" name="Constant_52978" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5560" name="Constant_338011" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
+		<layer id="3069" name="Constant_52980" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5561" name="Concat_338012" type="Concat" version="opset1">
+		<layer id="3070" name="Concat_52981" type="Concat" version="opset1">
 			<data axis="0" />
 			<input>
 				<port id="0" precision="I64">
@@ -85062,7 +46508,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5562" name="Broadcast_338015" type="Broadcast" version="opset3">
+		<layer id="3071" name="Broadcast_52984" type="Broadcast" version="opset3">
 			<data mode="numpy" />
 			<input>
 				<port id="0" precision="FP32" />
@@ -85079,8 +46525,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5563" name="ReadValue_330487" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.22.valuepresent.22.value" variable_type="f32" variable_shape="?,32,?,80" />
+		<layer id="3072" name="ReadValue_50223" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.18.keypresent.18.key" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -85090,7 +46536,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="past_key_values.22.value">
+				<port id="1" precision="FP32" names="past_key_values.18.key">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -85098,13 +46544,13 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5564" name="Constant_323520" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
+		<layer id="3073" name="Constant_47886" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="5565" name="Gather_323521" type="Gather" version="opset8">
+		<layer id="3074" name="Gather_47887" type="Gather" version="opset8">
 			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="FP32">
@@ -85127,41 +46573,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5566" name="Constant_277553" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3075" name="Constant_12074" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5567" name="Constant_277556" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3076" name="Constant_12076" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5568" name="Constant_277559" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3077" name="Constant_12078" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5569" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3078" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Slice_5" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -85170,17 +46606,20 @@
 					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3612">
+				<port id="5" precision="FP32" names="2416">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -85188,15 +46627,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5570" name="Constant_13033" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="3079" name="Constant_12087" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64" names="3613">
+				<port id="0" precision="I64" names="2417">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5571" name="__module.model.gpt_neox.layers.22.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
+		<layer id="3080" name="__module.model.gpt_neox.layers.18.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -85209,7 +46648,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3614">
+				<port id="2" precision="FP32" names="2418,key.73">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -85217,8 +46656,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5572" name="__module.model.gpt_neox.layers.22.attention/aten::cat/Concat_5" type="Concat" version="opset1">
-			<data axis="-2" />
+		<layer id="3081" name="Constant_12130" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3082" name="Constant_12132" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3083" name="Constant_12134" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3084" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Slice_6" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -85226,176 +46688,303 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="5" precision="FP32" names="2424,k.37">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3085" name="__module.model.gpt_neox.layers.18.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="present.22.value">
+				<port id="2" precision="FP32" names="2444">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5573" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Slice_11" type="Const" version="opset1">
-			<data element_type="boolean" shape="1, 1, 2048, 2048" offset="201244276" size="4194304" />
+		<layer id="3086" name="Constant_12411" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
-				<port id="0" precision="BOOL" names="3691">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5574" name="Constant_277563" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3087" name="Constant_12413" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>3</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5575" name="Constant_277562" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162298" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3088" name="Constant_12415" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5576" name="__module.model.gpt_neox.layers.22.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="3089" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Slice_7" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="5" precision="FP32" names="2454,x2.75">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5577" name="__module.model.gpt_neox.layers.22.attention/aten::size/Constant_8" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3090" name="Constant_37438" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5578" name="__module.model.gpt_neox.layers.22.attention/aten::size/Gather_8" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="3091" name="__module.model.gpt_neox.layers.18.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="3686" />
+				<port id="2" precision="FP32" names="2455">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5579" name="__module.model.gpt_neox.layers.22.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="3092" name="Constant_12308" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3093" name="Constant_12352" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3094" name="Constant_12354" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3095" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Slice_8" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="5" precision="FP32" names="2449">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5580" name="__module.model.gpt_neox.layers.22.attention/aten::size/Constant_9" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3096" name="__module.model.gpt_neox.layers.18.attention/aten::cat/Concat_2" type="Concat" version="opset1">
+			<data axis="-1" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I32" />
+				<port id="2" precision="FP32" names="2457">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5581" name="__module.model.gpt_neox.layers.22.attention/aten::size/Gather_9" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="3097" name="__module.model.gpt_neox.layers.18.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="3684" />
+				<port id="2" precision="FP32" names="2458">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5582" name="__module.model.gpt_neox.layers.22.attention/aten::sub/Subtract" type="Subtract" version="opset1">
+		<layer id="3098" name="__module.model.gpt_neox.layers.18.attention/aten::add/Add_1" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64" names="3688,3689" />
+				<port id="2" precision="FP32" names="2459">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5583" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Constant_4" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3099" name="Constant_12140" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5584" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="3100" name="Constant_12142" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5585" name="Constant_277561" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3101" name="Constant_12144" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5586" name="ScatterUpdate_277564" type="ScatterUpdate" version="opset3">
+		<layer id="3102" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Slice_9" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -85403,280 +46992,313 @@
 				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
 				<port id="4" precision="I64">
-					<dim>3</dim>
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="5587" name="Constant_277565" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="5" precision="FP32" names="2425">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5588" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1">
+		<layer id="3103" name="__module.model.gpt_neox.layers.18.attention/aten::cat/Concat_3" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="2463">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5589" name="ScatterUpdate_277566" type="ScatterUpdate" version="opset3">
+		<layer id="3104" name="__module.model.gpt_neox.layers.18.attention/aten::cat/Concat_4" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="present.18.key">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="3" precision="I32">
+			</output>
+		</layer>
+		<layer id="3105" name="Constant_52995" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
+			<output>
+				<port id="0" precision="FP32" />
+			</output>
+		</layer>
+		<layer id="3106" name="Constant_52988" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="3107" name="Constant_52990" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5590" name="Constant_277569" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438612" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3108" name="Constant_52992" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>3</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5591" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3109" name="Concat_52993" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>3</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>3</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>3</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="BOOL" names="3692">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
+				<port id="4" precision="I64">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5592" name="Constant_277642" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
+		<layer id="3110" name="Broadcast_52996" type="Broadcast" version="opset3">
+			<data mode="numpy" />
+			<input>
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="5593" name="Constant_277643" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5594" name="Constant_277639" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3111" name="ReadValue_50225" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.18.valuepresent.18.value" variable_type="f32" variable_shape="?,32,?,80" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="1" precision="FP32" names="past_key_values.18.value">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5595" name="Constant_277638" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3112" name="Constant_47889" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="5596" name="ScatterUpdate_277644" type="ScatterUpdate" version="opset3">
+		<layer id="3113" name="Gather_47890" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
+				<port id="1" precision="I32">
+					<dim>-1</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="2" precision="I64" />
+			</input>
+			<output>
+				<port id="3" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="3" precision="I32">
+			</output>
+		</layer>
+		<layer id="3114" name="Constant_12092" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="3115" name="Constant_12094" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5597" name="Constant_277647" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3116" name="Constant_12096" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5598" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3117" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Slice_10" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>2048</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="BOOL" names="3693,3694">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="5" precision="FP32" names="2419">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5599" name="__module.model.gpt_neox.layers.22.attention/aten::where/Convert_1_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="5600" name="__module.model.gpt_neox.layers.22.attention/aten::where/Convert_1" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="5601" name="__module.model.gpt_neox.layers.22.attention/aten::full/Convert" type="Const" version="opset1">
-			<data element_type="f32" shape="" offset="205438636" size="4" />
+		<layer id="3118" name="Constant_12105" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="FP32" names="3681" />
+				<port id="0" precision="I64" names="2420">
+					<dim>4</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5602" name="__module.model.gpt_neox.layers.22.attention/aten::where/Select" type="Select" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3119" name="__module.model.gpt_neox.layers.18.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
 			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
-				<port id="1" precision="FP32" />
-				<port id="2" precision="FP32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="3695,causal_mask.89">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="2421">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5603" name="Add_290470" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3120" name="__module.model.gpt_neox.layers.18.attention/aten::cat/Concat_5" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="present.18.value">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3121" name="Constant_12487" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5604" name="ShapeOf_290471" type="ShapeOf" version="opset3">
+		<layer id="3122" name="ShapeOf_39990" type="ShapeOf" version="opset3">
 			<data output_type="i64" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
@@ -85685,27 +47307,21 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5605" name="Constant_254659" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3123" name="Constant_39991" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116168" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5606" name="__module.model.gpt_neox.layers.22.attention/aten::size/Constant_10" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3124" name="Constant_39992" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="5607" name="__module.model.gpt_neox.layers.22.attention/aten::size/Gather_10" type="Gather" version="opset8">
+		<layer id="3125" name="Gather_39993" type="Gather" version="opset8">
 			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="I64">
@@ -85714,16 +47330,16 @@
 				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I32" />
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="3679">
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5608" name="__module.model.gpt_neox.layers.22.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="3126" name="Add_39996" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="I64">
 					<dim>1</dim>
@@ -85731,49 +47347,22 @@
 				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64" names="3696">
-					<dim>4</dim>
+				<port id="2" precision="I64" names="2468">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5609" name="__module.model.gpt_neox.layers.22.attention/aten::expand/Abs" type="Abs" version="opset1">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
+		<layer id="3127" name="Constant_12490" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5610" name="Maximum_290472" type="Maximum" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5611" name="__module.model.gpt_neox.layers.22.attention/aten::add/Add_2" type="Broadcast" version="opset3">
-			<data mode="numpy" />
+		<layer id="3128" name="__module.model.gpt_neox.layers.18.attention/aten::slice/Slice_14" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -85782,19 +47371,28 @@
 					<dim>-1</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3698">
-					<dim>-1</dim>
+				<port id="5" precision="FP32" names="2472">
 					<dim>-1</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
 					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5612" name="__module.model.gpt_neox.layers.22.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+		<layer id="3129" name="__module.model.gpt_neox.layers.18.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
 			<data causal="false" />
 			<input>
 				<port id="0" precision="FP32">
@@ -85817,13 +47415,13 @@
 				</port>
 				<port id="3" precision="FP32">
 					<dim>-1</dim>
-					<dim>-1</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
 					<dim>-1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3699,3700,sdpa_result.45,tensor.89">
+				<port id="4" precision="FP32" names="2473,attn_output.109">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -85831,15 +47429,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5613" name="Constant_13474" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="3130" name="__module.model.gpt_neox.layers.18.attention/aten::transpose/Constant" type="Const" version="opset1">
+			<data element_type="i32" shape="4" offset="201116200" size="16" />
 			<output>
-				<port id="0" precision="I64" names="3701">
+				<port id="0" precision="I32">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5614" name="__module.model.gpt_neox.layers.22.attention/aten::permute/Transpose_3" type="Transpose" version="opset1">
+		<layer id="3131" name="__module.model.gpt_neox.layers.18.attention/aten::transpose/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -85847,12 +47445,12 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="I32">
 					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3702">
+				<port id="2" precision="FP32" names="2474">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -85860,18 +47458,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5615" name="__module.model.gpt_neox.layers.22.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438650" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3132" name="Constant_41002" type="Const" version="opset1">
+			<data element_type="i64" shape="3" offset="201116216" size="24" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>3</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5616" name="__module.model.gpt_neox.layers.22.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
+		<layer id="3133" name="__module.model.gpt_neox.layers.18.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
 			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -85885,15 +47480,15 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3707">
+				<port id="2" precision="FP32" names="2477">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5617" name="self.model.gpt_neox.layers.22.attention.dense.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 2560" offset="1938574034" size="6553600" />
+		<layer id="3134" name="self.model.gpt_neox.layers.18.attention.dense.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 2560" offset="1619873360" size="6553600" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>2560</dim>
@@ -85901,7 +47496,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5618" name="Convert_22314" type="Convert" version="opset1">
+		<layer id="3135" name="Convert_75374" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -85916,8 +47511,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5619" name="self.model.gpt_neox.layers.22.attention.dense.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="1945127634" size="2560" />
+		<layer id="3136" name="self.model.gpt_neox.layers.18.attention.dense.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="1626426960" size="2560" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>2560</dim>
@@ -85925,7 +47520,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5620" name="Convert_22317" type="Convert" version="opset1">
+		<layer id="3137" name="Convert_75377" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -85940,7 +47535,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5621" name="self.model.gpt_neox.layers.22.attention.dense.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="3138" name="self.model.gpt_neox.layers.18.attention.dense.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -85959,8 +47554,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5622" name="self.model.gpt_neox.layers.22.attention.dense.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="1945130194" size="5120" />
+		<layer id="3139" name="self.model.gpt_neox.layers.18.attention.dense.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="1626429520" size="5120" />
 			<output>
 				<port id="0" precision="FP16">
 					<dim>2560</dim>
@@ -85968,7 +47563,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5623" name="self.model.gpt_neox.layers.22.attention.dense.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="3140" name="self.model.gpt_neox.layers.18.attention.dense.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -85987,11 +47582,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5624" name="self.model.gpt_neox.layers.22.attention.dense.weight" type="Convert" version="opset1">
+		<layer id="3141" name="self.model.gpt_neox.layers.18.attention.dense.weight/fq_weights_1/convert" type="Convert" version="opset1">
 			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
 			<input>
 				<port id="0" precision="FP16">
 					<dim>2560</dim>
@@ -86005,7 +47597,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5625" name="__module.model.gpt_neox.layers.22.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="3142" name="__module.model.gpt_neox.layers.18.attention.dense/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -86026,37 +47618,121 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5626" name="Constant_299298_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1945135314" size="5120" />
+		<layer id="3143" name="Constant_37439" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1626434640" size="10240" />
 			<output>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5627" name="Constant_299298" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="3144" name="__module.model.gpt_neox.layers.18.attention.dense/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
+				<port id="2" precision="FP32" names="2480,2482,input.73">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3145" name="__module.model.gpt_neox.layers.18/aten::add/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="2500">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3146" name="__module.model.gpt_neox.layers.18/aten::add/Add_1" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
 				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="2501,2503,hidden_states.79">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3147" name="__module.model.gpt_neox.layers.19.input_layernorm/aten::layer_norm/Multiply" type="Const" version="opset1">
+			<data element_type="i32" shape="1" offset="128867668" size="4" />
+			<output>
+				<port id="0" precision="I32">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3148" name="__module.model.gpt_neox.layers.19.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
+			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="I32">
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3149" name="Constant_37440" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1626444880" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5628" name="__module.model.gpt_neox.layers.22.attention.dense/aten::linear/Add" type="Add" version="opset1">
+		<layer id="3150" name="__module.model.gpt_neox.layers.19.post_attention_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -86071,14 +47747,24 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3710,3712,input.91">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5629" name="__module.model.gpt_neox.layers.22/aten::add/Add" type="Add" version="opset1">
+		<layer id="3151" name="Constant_37441" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1626455120" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3152" name="__module.model.gpt_neox.layers.19.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -86087,59 +47773,338 @@
 					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="2601">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
+			</output>
+		</layer>
+		<layer id="3153" name="self.model.gpt_neox.layers.19.mlp.dense_h_to_4h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 2560" offset="1626465360" size="26214400" />
+			<output>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3154" name="Convert_74564" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3155" name="self.model.gpt_neox.layers.19.mlp.dense_h_to_4h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 1" offset="1652679760" size="10240" />
+			<output>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3156" name="Convert_74567" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3157" name="self.model.gpt_neox.layers.19.mlp.dense_h_to_4h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3158" name="self.model.gpt_neox.layers.19.mlp.dense_h_to_4h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="10240, 1" offset="1652690000" size="20480" />
+			<output>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3159" name="self.model.gpt_neox.layers.19.mlp.dense_h_to_4h.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3160" name="__module.model.gpt_neox.layers.19.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<rt_info>
+				<attribute name="decompression" version="0" />
+			</rt_info>
+			<input>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3730">
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3161" name="__module.model.gpt_neox.layers.19.mlp.dense_h_to_4h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
+			<input>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3162" name="self.model.gpt_neox.layers.19.mlp.dense_h_to_4h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="10240" offset="1652710480" size="20480" />
+			<output>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.19.mlp.dense_h_to_4h.bias">
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3163" name="__module.model.gpt_neox.layers.19.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<rt_info>
+				<attribute name="decompression" version="0" />
+			</rt_info>
+			<input>
+				<port id="0" precision="BF16">
+					<dim>10240</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5630" name="__module.model.gpt_neox.layers.22/aten::add/Add_1" type="Add" version="opset1">
+		<layer id="3164" name="__module.model.gpt_neox.layers.19.mlp.dense_h_to_4h/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 				<port id="1" precision="FP32">
+					<dim>10240</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="2606">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3165" name="__module.model.gpt_neox.layers.19.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
+			<data approximation_mode="ERF" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3731,3733,hidden_states.47">
+				<port id="1" precision="FP32" names="2607">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3166" name="self.model.gpt_neox.layers.19.mlp.dense_4h_to_h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 10240" offset="1652730960" size="26214400" />
+			<output>
+				<port id="0" precision="U8">
 					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5631" name="__module.model.gpt_neox.layers.23.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867670" size="4" />
+		<layer id="3167" name="Convert_74574" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I32">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3168" name="self.model.gpt_neox.layers.19.mlp.dense_4h_to_h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="1678945360" size="2560" />
+			<output>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3169" name="Convert_74577" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3170" name="self.model.gpt_neox.layers.19.mlp.dense_4h_to_h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3171" name="self.model.gpt_neox.layers.19.mlp.dense_4h_to_h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="1678947920" size="5120" />
+			<output>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3172" name="self.model.gpt_neox.layers.19.mlp.dense_4h_to_h.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3173" name="__module.model.gpt_neox.layers.19.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<rt_info>
+				<attribute name="decompression" version="0" />
+			</rt_info>
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5632" name="__module.model.gpt_neox.layers.23.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
-			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
+		<layer id="3174" name="__module.model.gpt_neox.layers.19.mlp.dense_4h_to_h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="1" precision="I32">
-					<dim>1</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
@@ -86150,37 +48115,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5633" name="Constant_299299_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1945140434" size="5120" />
+		<layer id="3175" name="self.model.gpt_neox.layers.19.mlp.dense_4h_to_h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="2560" offset="1678953040" size="5120" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.19.mlp.dense_4h_to_h.bias">
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5634" name="Constant_299299" type="Convert" version="opset1">
+		<layer id="3176" name="__module.model.gpt_neox.layers.19.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
 			</rt_info>
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16">
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5635" name="__module.model.gpt_neox.layers.23.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
+		<layer id="3177" name="__module.model.gpt_neox.layers.19.mlp.dense_4h_to_h/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -86189,50 +48148,60 @@
 					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="2610,input.79">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5636" name="Constant_299300_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1945145554" size="5120" />
+		<layer id="3178" name="Constant_37442" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1678958160" size="10240" />
 			<output>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5637" name="Constant_299300" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="3179" name="__module.model.gpt_neox.layers.19.input_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3180" name="Constant_37443" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1678968400" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5638" name="__module.model.gpt_neox.layers.23.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+		<layer id="3181" name="__module.model.gpt_neox.layers.19.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -86247,127 +48216,124 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3868">
+				<port id="2" precision="FP32" names="2513,hidden_states.77">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5639" name="self.model.gpt_neox.layers.23.mlp.dense_h_to_4h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 2560" offset="1945150674" size="26214400" />
+		<layer id="3182" name="self.model.gpt_neox.layers.19.attention.query_key_value.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 2560" offset="1678978640" size="19660800" />
 			<output>
 				<port id="0" precision="U8">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5640" name="Convert_22364" type="Convert" version="opset1">
+		<layer id="3183" name="Convert_75033" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP16">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5641" name="self.model.gpt_neox.layers.23.mlp.dense_h_to_4h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 1" offset="1971365074" size="10240" />
+		<layer id="3184" name="self.model.gpt_neox.layers.19.attention.query_key_value.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 1" offset="1698639440" size="7680" />
 			<output>
 				<port id="0" precision="U8">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5642" name="Convert_22367" type="Convert" version="opset1">
+		<layer id="3185" name="Convert_75036" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP16">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5643" name="self.model.gpt_neox.layers.23.mlp.dense_h_to_4h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="3186" name="self.model.gpt_neox.layers.19.attention.query_key_value.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP16">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
 				<port id="2" precision="FP16">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5644" name="self.model.gpt_neox.layers.23.mlp.dense_h_to_4h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="10240, 1" offset="1971375314" size="20480" />
+		<layer id="3187" name="self.model.gpt_neox.layers.19.attention.query_key_value.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="7680, 1" offset="1698647120" size="15360" />
 			<output>
 				<port id="0" precision="FP16">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5645" name="self.model.gpt_neox.layers.23.mlp.dense_h_to_4h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="3188" name="self.model.gpt_neox.layers.19.attention.query_key_value.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP16">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
 				<port id="2" precision="FP16">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5646" name="self.model.gpt_neox.layers.23.mlp.dense_h_to_4h.weight" type="Convert" version="opset1">
+		<layer id="3189" name="self.model.gpt_neox.layers.19.attention.query_key_value.weight/fq_weights_1/convert" type="Convert" version="opset1">
 			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
 			<input>
 				<port id="0" precision="FP16">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP32">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5647" name="__module.model.gpt_neox.layers.23.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="3190" name="__module.model.gpt_neox.layers.19.attention.query_key_value/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -86376,7 +48342,7 @@
 					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
@@ -86384,859 +48350,1101 @@
 				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5648" name="Constant_299301_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 10240" offset="1971395794" size="20480" />
+		<layer id="3191" name="Constant_37444" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 7680" offset="1698662480" size="30720" />
 			<output>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>10240</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5649" name="Constant_299301" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="3192" name="__module.model.gpt_neox.layers.19.attention.query_key_value/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>7680</dim>
+				</port>
+				<port id="1" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>10240</dim>
+					<dim>7680</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
+				<port id="2" precision="FP32" names="2520,qkv.77">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>7680</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3193" name="__module.model.gpt_neox.layers.19.attention/prim::ListConstruct" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201115992" size="32" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3194" name="__module.model.gpt_neox.layers.19.attention/aten::view/Reshape" type="Reshape" version="opset1">
+			<data special_zero="true" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>7680</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>4</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="2524,qkv.79">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3195" name="Constant_12686" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3196" name="Constant_12688" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5650" name="__module.model.gpt_neox.layers.23.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3197" name="Constant_12690" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3198" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Slice" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
-				<port id="1" precision="FP32">
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3873">
+				<port id="5" precision="FP32" names="2525">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5651" name="__module.model.gpt_neox.layers.23.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
-			<data approximation_mode="ERF" />
+		<layer id="3199" name="Constant_12699" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
+			<output>
+				<port id="0" precision="I64" names="2526">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3200" name="__module.model.gpt_neox.layers.19.attention/aten::permute/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="3874">
+				<port id="2" precision="FP32" names="2527,query.39">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5652" name="self.model.gpt_neox.layers.23.mlp.dense_4h_to_h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 10240" offset="1971416274" size="26214400" />
+		<layer id="3201" name="Constant_12740" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5653" name="Convert_22374" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
+		<layer id="3202" name="Constant_12742" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5654" name="self.model.gpt_neox.layers.23.mlp.dense_4h_to_h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="1997630674" size="2560" />
+		<layer id="3203" name="Constant_12744" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5655" name="Convert_22377" type="Convert" version="opset1">
-			<data destination_type="f16" />
+		<layer id="3204" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Slice_1" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
+				<port id="5" precision="FP32" names="2534,q.39">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5656" name="self.model.gpt_neox.layers.23.mlp.dense_4h_to_h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="3205" name="__module.model.gpt_neox.layers.19.attention/aten::mul/Multiply" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
 					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="2" precision="FP32" names="2540">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5657" name="self.model.gpt_neox.layers.23.mlp.dense_4h_to_h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="1997633234" size="5120" />
+		<layer id="3206" name="Constant_12898" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5658" name="self.model.gpt_neox.layers.23.mlp.dense_4h_to_h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+		<layer id="3207" name="Constant_12900" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="3208" name="Constant_12902" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5659" name="self.model.gpt_neox.layers.23.mlp.dense_4h_to_h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="3209" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Slice_2" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="5" precision="FP32" names="2550,x2.77">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5660" name="__module.model.gpt_neox.layers.23.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
+		<layer id="3210" name="Constant_37445" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3211" name="__module.model.gpt_neox.layers.19.attention/aten::neg/Multiply" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>10</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="2551">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5661" name="Constant_299302_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1997638354" size="5120" />
+		<layer id="3212" name="Constant_12795" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="FP16">
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3213" name="Constant_12839" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5662" name="Constant_299302" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="3214" name="Constant_12841" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3215" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Slice_3" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
+				<port id="5" precision="FP32" names="2545">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3216" name="__module.model.gpt_neox.layers.19.attention/aten::cat/Concat" type="Concat" version="opset1">
+			<data axis="-1" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="2553">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5663" name="__module.model.gpt_neox.layers.23.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1">
+		<layer id="3217" name="__module.model.gpt_neox.layers.19.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
+					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3877,input.97">
+				<port id="2" precision="FP32" names="2554">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5664" name="Constant_299303_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1997643474" size="5120" />
+		<layer id="3218" name="__module.model.gpt_neox.layers.19.attention/aten::add/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="FP16">
+				<port id="2" precision="FP32" names="2555">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3219" name="Constant_12750" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3220" name="Constant_12752" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5665" name="Constant_299303" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="3221" name="Constant_12754" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3222" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Slice_4" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+				<port id="5" precision="FP32" names="2535">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5666" name="__module.model.gpt_neox.layers.23.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3223" name="__module.model.gpt_neox.layers.19.attention/aten::cat/Concat_1" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="2573">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3224" name="Constant_53007" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
+			<output>
+				<port id="0" precision="FP32" />
+			</output>
+		</layer>
+		<layer id="3225" name="Constant_53000" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3226" name="Constant_53002" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3227" name="Constant_53004" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3228" name="Concat_53005" type="Concat" version="opset1">
+			<data axis="0" />
+			<input>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="FP32">
+				<port id="2" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5667" name="Constant_299304_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="1997648594" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+				<port id="4" precision="I64">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5668" name="Constant_299304" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="3229" name="Broadcast_53008" type="Broadcast" version="opset3">
+			<data mode="numpy" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5669" name="__module.model.gpt_neox.layers.23.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3230" name="ReadValue_50227" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.19.keypresent.19.key" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3743">
+				<port id="1" precision="FP32" names="past_key_values.19.key">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5670" name="self.model.gpt_neox.layers.23.attention.query_key_value.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 2560" offset="1997653714" size="19660800" />
+		<layer id="3231" name="Constant_47892" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="5671" name="Convert_22344" type="Convert" version="opset1">
-			<data destination_type="f16" />
+		<layer id="3232" name="Gather_47893" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
+				<port id="1" precision="I32">
+					<dim>-1</dim>
+				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="3" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5672" name="self.model.gpt_neox.layers.23.attention.query_key_value.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 1" offset="2017314514" size="7680" />
+		<layer id="3233" name="Constant_12704" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5673" name="Convert_22347" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
+		<layer id="3234" name="Constant_12706" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="3235" name="Constant_12708" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5674" name="self.model.gpt_neox.layers.23.attention.query_key_value.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3236" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Slice_5" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
+				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5675" name="self.model.gpt_neox.layers.23.attention.query_key_value.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="7680, 1" offset="2017322194" size="15360" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
+				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="5676" name="self.model.gpt_neox.layers.23.attention.query_key_value.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="3" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="5" precision="FP32" names="2528">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5677" name="self.model.gpt_neox.layers.23.attention.query_key_value.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="3237" name="Constant_12717" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64" names="2529">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5678" name="__module.model.gpt_neox.layers.23.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
+		<layer id="3238" name="__module.model.gpt_neox.layers.19.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="2530,key.77">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5679" name="Constant_299305_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 7680" offset="2017337554" size="15360" />
+		<layer id="3239" name="Constant_12760" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5680" name="Constant_299305" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+		<layer id="3240" name="Constant_12762" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>7680</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="3241" name="Constant_12764" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5681" name="__module.model.gpt_neox.layers.23.attention.query_key_value/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3242" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Slice_6" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>7680</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3750,qkv.93">
+				<port id="5" precision="FP32" names="2536,k.39">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5682" name="__module.model.gpt_neox.layers.23.attention/prim::ListConstruct/Concat" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080154" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5683" name="__module.model.gpt_neox.layers.23.attention/aten::view/Reshape" type="Reshape" version="opset1">
-			<data special_zero="true" />
+		<layer id="3243" name="__module.model.gpt_neox.layers.19.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3754,qkv.95">
-					<dim>-1</dim>
+				<port id="2" precision="FP32" names="2556">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>240</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5684" name="Constant_277654" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3244" name="Constant_13041" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5685" name="Constant_277657" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3245" name="Constant_13043" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5686" name="Constant_277660" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3246" name="Constant_13045" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5687" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3247" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Slice_7" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>240</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3755">
-					<dim>-1</dim>
+				<port id="5" precision="FP32" names="2566,x2.79">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5688" name="Constant_13586" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="3248" name="Constant_37446" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="0" precision="I64" names="3756">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5689" name="__module.model.gpt_neox.layers.23.attention/aten::permute/Transpose" type="Transpose" version="opset1">
+		<layer id="3249" name="__module.model.gpt_neox.layers.19.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3757,query.139">
+				<port id="2" precision="FP32" names="2567">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5690" name="Constant_277666" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3250" name="Constant_12938" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5691" name="Constant_277669" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3251" name="Constant_12982" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5692" name="Constant_277672" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3252" name="Constant_12984" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5693" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3253" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Slice_8" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3764,q.47">
+				<port id="5" precision="FP32" names="2561">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5694" name="self.model.gpt_neox.layers.23.attention.rotary_emb.cos_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201080346" size="81920" />
+		<layer id="3254" name="__module.model.gpt_neox.layers.19.attention/aten::cat/Concat_2" type="Concat" version="opset1">
+			<data axis="-1" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.23.attention.rotary_emb.cos_cached">
-					<dim>2048</dim>
+				<port id="2" precision="FP32" names="2569">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5695" name="self.model.gpt_neox.layers.23.attention.rotary_emb.cos_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="3255" name="__module.model.gpt_neox.layers.19.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
+				<port id="2" precision="FP32" names="2570">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5696" name="__module.model.gpt_neox.layers.23.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3256" name="__module.model.gpt_neox.layers.19.attention/aten::add/Add_1" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="2571">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5697" name="Constant_277678" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3257" name="Constant_12770" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5698" name="Constant_277681" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3258" name="Constant_12772" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5699" name="Constant_277684" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3259" name="Constant_12774" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5700" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3260" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Slice_9" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>240</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3758">
-					<dim>-1</dim>
+				<port id="5" precision="FP32" names="2537">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5701" name="Constant_13600" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="3759">
-					<dim>4</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5702" name="__module.model.gpt_neox.layers.23.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
+		<layer id="3261" name="__module.model.gpt_neox.layers.19.attention/aten::cat/Concat_3" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3760,key.139">
+				<port id="2" precision="FP32" names="2575">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -87244,8 +49452,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5703" name="__module.model.gpt_neox.layers.23.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="3262" name="__module.model.gpt_neox.layers.19.attention/aten::cat/Concat_4" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -87253,78 +49461,53 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5704" name="__module.model.gpt_neox.layers.23.attention/aten::size/Constant_2" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="5705" name="__module.model.gpt_neox.layers.23.attention/aten::size/Gather_2" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="3768" />
-			</output>
-		</layer>
-		<layer id="5706" name="Constant_338027_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
+				<port id="2" precision="FP32" names="present.19.key">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5707" name="Constant_338027" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
+		<layer id="3263" name="Constant_53019" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="1" precision="FP32" />
+				<port id="0" precision="FP32" />
 			</output>
 		</layer>
-		<layer id="5708" name="Constant_338020" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
+		<layer id="3264" name="Constant_53012" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5709" name="Constant_338022" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
+		<layer id="3265" name="Constant_53014" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5710" name="Constant_338024" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
+		<layer id="3266" name="Constant_53016" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5711" name="Concat_338025" type="Concat" version="opset1">
+		<layer id="3267" name="Concat_53017" type="Concat" version="opset1">
 			<data axis="0" />
 			<input>
 				<port id="0" precision="I64">
@@ -87346,7 +49529,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5712" name="Broadcast_338028" type="Broadcast" version="opset3">
+		<layer id="3268" name="Broadcast_53020" type="Broadcast" version="opset3">
 			<data mode="numpy" />
 			<input>
 				<port id="0" precision="FP32" />
@@ -87363,8 +49546,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5713" name="ReadValue_330489" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.23.keypresent.23.key" variable_type="f32" variable_shape="?,32,?,80" />
+		<layer id="3269" name="ReadValue_50229" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.19.valuepresent.19.value" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -87374,7 +49557,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="past_key_values.23.key">
+				<port id="1" precision="FP32" names="past_key_values.19.value">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -87382,13 +49565,13 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5714" name="Constant_323523" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
+		<layer id="3270" name="Constant_47895" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="5715" name="Gather_323524" type="Gather" version="opset8">
+		<layer id="3271" name="Gather_47896" type="Gather" version="opset8">
 			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="FP32">
@@ -87396,106 +49579,52 @@
 					<dim>32</dim>
 					<dim>-1</dim>
 					<dim>80</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I64" />
-			</input>
-			<output>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5716" name="__module.model.gpt_neox.layers.23.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5717" name="__module.model.gpt_neox.layers.23.attention/aten::size/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="5718" name="__module.model.gpt_neox.layers.23.attention/aten::size/Gather_3" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="3770" />
-			</output>
-		</layer>
-		<layer id="5719" name="__module.model.gpt_neox.layers.23.attention/aten::add_/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
+				</port>
+				<port id="1" precision="I32">
+					<dim>-1</dim>
+				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="I64" names="3769,3773,3774,seq_len.93" />
+				<port id="3" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5720" name="__module.model.gpt_neox.layers.23.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3272" name="Constant_12722" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5721" name="__module.model.gpt_neox.layers.23.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="3273" name="Constant_12724" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5722" name="__module.model.gpt_neox.layers.23.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3274" name="Constant_12726" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5723" name="__module.model.gpt_neox.layers.23.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3275" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Slice_10" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -87506,114 +49635,89 @@
 				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3777,3780,cos.93">
+				<port id="5" precision="FP32" names="2531">
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5724" name="Constant_38335" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="5725" name="__module.model.gpt_neox.layers.23.attention/aten::index/Gather" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
+		<layer id="3276" name="Constant_12735" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="3" precision="FP32" names="3783">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="I64" names="2532">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5726" name="__module.model.gpt_neox.layers.23.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1">
+		<layer id="3277" name="__module.model.gpt_neox.layers.19.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
-				<port id="1" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3784,cos.95">
+				<port id="2" precision="FP32" names="2533">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5727" name="__module.model.gpt_neox.layers.23.attention/aten::mul/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3278" name="__module.model.gpt_neox.layers.19.attention/aten::cat/Concat_5" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3788">
+				<port id="2" precision="FP32" names="present.19.value">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5728" name="Constant_277689" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5729" name="Constant_277688" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3279" name="Constant_13117" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5730" name="__module.model.gpt_neox.layers.23.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3">
+		<layer id="3280" name="ShapeOf_40061" type="ShapeOf" version="opset3">
 			<data output_type="i64" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
@@ -87622,1259 +49726,1139 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5731" name="__module.model.gpt_neox.layers.23.attention/aten::size/Constant_5" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3281" name="Constant_40062" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116168" size="8" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="5732" name="__module.model.gpt_neox.layers.23.attention/aten::size/Gather_5" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="3789,3794" />
 			</output>
 		</layer>
-		<layer id="5733" name="__module.model.gpt_neox.layers.23.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
+		<layer id="3282" name="Constant_40063" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="2" precision="I64" />
-			</output>
-		</layer>
-		<layer id="5734" name="__module.model.gpt_neox.layers.23.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1">
-			<input>
 				<port id="0" precision="I64" />
-			</input>
-			<output>
-				<port id="1" precision="I64" names="3791,3792,3796,3797" />
-			</output>
-		</layer>
-		<layer id="5735" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Constant_1" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
 			</output>
 		</layer>
-		<layer id="5736" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1">
+		<layer id="3283" name="Gather_40064" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="I64">
+					<dim>4</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="5737" name="Constant_277687" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+				<port id="2" precision="I64" />
+			</input>
 			<output>
-				<port id="0" precision="I32">
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5738" name="ScatterUpdate_277690" type="ScatterUpdate" version="opset3">
+		<layer id="3284" name="Add_40067" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5739" name="Constant_277693" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="I64" names="2580">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5740" name="Constant_277696" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3285" name="Constant_13120" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5741" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3286" name="__module.model.gpt_neox.layers.19.attention/aten::slice/Slice_14" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="3798,x2.93">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5742" name="Constant_299306_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="5743" name="Constant_299306" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="5" precision="FP32" names="2584">
+					<dim>-1</dim>
 					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5744" name="__module.model.gpt_neox.layers.23.attention/aten::neg/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3287" name="__module.model.gpt_neox.layers.19.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+			<data causal="false" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="3" precision="FP32">
+					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3799">
+				<port id="4" precision="FP32" names="2585,attn_output.115">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5745" name="Constant_277741" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5746" name="Constant_277746" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3288" name="__module.model.gpt_neox.layers.19.attention/aten::transpose/Constant" type="Const" version="opset1">
+			<data element_type="i32" shape="4" offset="201116200" size="16" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="I32">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5747" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3289" name="__module.model.gpt_neox.layers.19.attention/aten::transpose/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>80</dim>
 				</port>
-				<port id="3" precision="I64">
+				<port id="1" precision="I32">
 					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3793">
+				<port id="2" precision="FP32" names="2586">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5748" name="__module.model.gpt_neox.layers.23.attention/aten::cat/Concat" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="3290" name="Constant_41003" type="Const" version="opset1">
+			<data element_type="i64" shape="3" offset="201116216" size="24" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>3</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3291" name="__module.model.gpt_neox.layers.19.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
+			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>3</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3801">
+				<port id="2" precision="FP32" names="2589">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5749" name="self.model.gpt_neox.layers.23.attention.rotary_emb.sin_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201162356" size="81920" />
+		<layer id="3292" name="self.model.gpt_neox.layers.19.attention.dense.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 2560" offset="1698693200" size="6553600" />
 			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.23.attention.rotary_emb.sin_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5750" name="self.model.gpt_neox.layers.23.attention.rotary_emb.sin_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="3293" name="Convert_75385" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5751" name="__module.model.gpt_neox.layers.23.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3294" name="self.model.gpt_neox.layers.19.attention.dense.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="1705246800" size="2560" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="U8">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5752" name="__module.model.gpt_neox.layers.23.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3295" name="Convert_75388" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5753" name="__module.model.gpt_neox.layers.23.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3296" name="self.model.gpt_neox.layers.19.attention.dense.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3778,3781,sin.93">
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5754" name="Constant_38338" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
+		<layer id="3297" name="self.model.gpt_neox.layers.19.attention.dense.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="1705249360" size="5120" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5755" name="__module.model.gpt_neox.layers.23.attention/aten::index/Gather_1" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="3298" name="self.model.gpt_neox.layers.19.attention.dense.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
 				</port>
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="3786">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5756" name="__module.model.gpt_neox.layers.23.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1">
+		<layer id="3299" name="self.model.gpt_neox.layers.19.attention.dense.weight/fq_weights_1/convert" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3787,sin.95">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5757" name="__module.model.gpt_neox.layers.23.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3300" name="__module.model.gpt_neox.layers.19.attention.dense/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3802">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3301" name="Constant_37447" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1705254480" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5758" name="__module.model.gpt_neox.layers.23.attention/aten::add/Add" type="Add" version="opset1">
+		<layer id="3302" name="__module.model.gpt_neox.layers.19.attention.dense/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3803">
+				<port id="2" precision="FP32" names="2592,2594,input.77">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5759" name="Constant_277792" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5760" name="Constant_277795" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5761" name="Constant_277798" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5762" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3303" name="__module.model.gpt_neox.layers.19/aten::add/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3765">
+				<port id="2" precision="FP32" names="2612">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>60</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5763" name="__module.model.gpt_neox.layers.23.attention/aten::cat/Concat_1" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="3304" name="__module.model.gpt_neox.layers.19/aten::add/Add_1" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>60</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3821,3831,query.141,query.143">
+				<port id="2" precision="FP32" names="2613,2615,hidden_states.83">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5764" name="Constant_277804" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5765" name="Constant_277807" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5766" name="Constant_277810" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3305" name="__module.model.gpt_neox.layers.20.input_layernorm/aten::layer_norm/Multiply" type="Const" version="opset1">
+			<data element_type="i32" shape="1" offset="128867668" size="4" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="I32">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5767" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3306" name="__module.model.gpt_neox.layers.20.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
+			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="I32">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3766,k.47">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3307" name="Constant_37448" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1705264720" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5768" name="__module.model.gpt_neox.layers.23.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
+		<layer id="3308" name="__module.model.gpt_neox.layers.20.post_attention_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3804">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5769" name="Constant_277814" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5770" name="Constant_277813" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3309" name="Constant_37449" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1705274960" size="10240" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="FP32">
+					<dim>1</dim>
 					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5771" name="__module.model.gpt_neox.layers.23.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="3310" name="__module.model.gpt_neox.layers.20.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="2713">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5772" name="__module.model.gpt_neox.layers.23.attention/aten::size/Constant_7" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3311" name="self.model.gpt_neox.layers.20.mlp.dense_h_to_4h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 2560" offset="1705285200" size="26214400" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5773" name="__module.model.gpt_neox.layers.23.attention/aten::size/Gather_7" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="3312" name="Convert_74584" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="3805,3810" />
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5774" name="__module.model.gpt_neox.layers.23.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
+		<layer id="3313" name="self.model.gpt_neox.layers.20.mlp.dense_h_to_4h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 1" offset="1731499600" size="10240" />
 			<output>
-				<port id="2" precision="I64" />
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5775" name="__module.model.gpt_neox.layers.23.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1">
+		<layer id="3314" name="Convert_74587" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64" />
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64" names="3807,3808,3812,3813" />
-			</output>
-		</layer>
-		<layer id="5776" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5777" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1">
+		<layer id="3315" name="self.model.gpt_neox.layers.20.mlp.dense_h_to_4h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5778" name="Constant_277812" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3316" name="self.model.gpt_neox.layers.20.mlp.dense_h_to_4h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="10240, 1" offset="1731509840" size="20480" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5779" name="ScatterUpdate_277815" type="ScatterUpdate" version="opset3">
+		<layer id="3317" name="self.model.gpt_neox.layers.20.mlp.dense_h_to_4h.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5780" name="Constant_277818" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
+		<layer id="3318" name="__module.model.gpt_neox.layers.20.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<rt_info>
-				<attribute name="precise" version="0" />
+				<attribute name="decompression" version="0" />
 			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+			<input>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="5781" name="Constant_277821" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5782" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3319" name="__module.model.gpt_neox.layers.20.mlp.dense_h_to_4h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3814,x2.95">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5783" name="Constant_299307_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
+		<layer id="3320" name="self.model.gpt_neox.layers.20.mlp.dense_h_to_4h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="10240" offset="1731530320" size="20480" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.20.mlp.dense_h_to_4h.bias">
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5784" name="Constant_299307" type="Convert" version="opset1">
+		<layer id="3321" name="__module.model.gpt_neox.layers.20.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
 			</rt_info>
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16">
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5785" name="__module.model.gpt_neox.layers.23.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
+		<layer id="3322" name="__module.model.gpt_neox.layers.20.mlp.dense_h_to_4h/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>10240</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3815">
+				<port id="2" precision="FP32" names="2718">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5786" name="Constant_277866" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3323" name="__module.model.gpt_neox.layers.20.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
+			<data approximation_mode="ERF" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32" names="2719">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5787" name="Constant_277871" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3324" name="self.model.gpt_neox.layers.20.mlp.dense_4h_to_h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 10240" offset="1731550800" size="26214400" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5788" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3325" name="Convert_74594" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+			</input>
+			<output>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+			</output>
+		</layer>
+		<layer id="3326" name="self.model.gpt_neox.layers.20.mlp.dense_4h_to_h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="1757765200" size="2560" />
+			<output>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+			</output>
+		</layer>
+		<layer id="3327" name="Convert_74597" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3809">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5789" name="__module.model.gpt_neox.layers.23.attention/aten::cat/Concat_2" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="3328" name="self.model.gpt_neox.layers.20.mlp.dense_4h_to_h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3817">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3329" name="self.model.gpt_neox.layers.20.mlp.dense_4h_to_h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="1757767760" size="5120" />
+			<output>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5790" name="__module.model.gpt_neox.layers.23.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
+		<layer id="3330" name="self.model.gpt_neox.layers.20.mlp.dense_4h_to_h.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3818">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5791" name="__module.model.gpt_neox.layers.23.attention/aten::add/Add_1" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3331" name="__module.model.gpt_neox.layers.20.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<rt_info>
+				<attribute name="decompression" version="0" />
+			</rt_info>
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3332" name="__module.model.gpt_neox.layers.20.mlp.dense_4h_to_h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>10240</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3819">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5792" name="Constant_277916" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3333" name="self.model.gpt_neox.layers.20.mlp.dense_4h_to_h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="2560" offset="1757772880" size="5120" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.20.mlp.dense_4h_to_h.bias">
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5793" name="Constant_277919" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
+		<layer id="3334" name="__module.model.gpt_neox.layers.20.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<rt_info>
-				<attribute name="precise" version="0" />
+				<attribute name="decompression" version="0" />
 			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+			<input>
+				<port id="0" precision="BF16">
+					<dim>2560</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="5794" name="Constant_277922" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5795" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3335" name="__module.model.gpt_neox.layers.20.mlp.dense_4h_to_h/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3767">
+				<port id="2" precision="FP32" names="2722,input.83">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>60</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5796" name="__module.model.gpt_neox.layers.23.attention/aten::cat/Concat_3" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="3336" name="Constant_37450" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1757778000" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3337" name="__module.model.gpt_neox.layers.20.input_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3823">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5797" name="__module.model.gpt_neox.layers.23.attention/aten::to/Convert_1" type="Concat" version="opset1">
-			<data axis="-2" />
+		<layer id="3338" name="Constant_37451" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1757788240" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3339" name="__module.model.gpt_neox.layers.20.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="present.23.key">
+				<port id="2" precision="FP32" names="2625,hidden_states.81">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5798" name="Constant_338040_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
+		<layer id="3340" name="self.model.gpt_neox.layers.20.attention.query_key_value.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 2560" offset="1757798480" size="19660800" />
 			<output>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5799" name="Constant_338040" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="3341" name="Convert_75044" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" />
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5800" name="Constant_338033" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
+		<layer id="3342" name="self.model.gpt_neox.layers.20.attention.query_key_value.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 1" offset="1777459280" size="7680" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="U8">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5801" name="Constant_338035" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
+		<layer id="3343" name="Convert_75047" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>1</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5802" name="Constant_338037" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
+		<layer id="3344" name="self.model.gpt_neox.layers.20.attention.query_key_value.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
+					<dim>1</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
+				<port id="2" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3345" name="self.model.gpt_neox.layers.20.attention.query_key_value.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="7680, 1" offset="1777466960" size="15360" />
+			<output>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5803" name="Concat_338038" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="3346" name="self.model.gpt_neox.layers.20.attention.query_key_value.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
+			</output>
+		</layer>
+		<layer id="3347" name="self.model.gpt_neox.layers.20.attention.query_key_value.weight/fq_weights_1/convert" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5804" name="Broadcast_338041" type="Broadcast" version="opset3">
-			<data mode="numpy" />
+		<layer id="3348" name="__module.model.gpt_neox.layers.20.attention.query_key_value/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5805" name="ReadValue_330491" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.23.valuepresent.23.value" variable_type="f32" variable_shape="?,32,?,80" />
+		<layer id="3349" name="Constant_37452" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 7680" offset="1777482320" size="30720" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>7680</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3350" name="__module.model.gpt_neox.layers.20.attention.query_key_value/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>7680</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>7680</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="past_key_values.23.value">
+				<port id="2" precision="FP32" names="2632,qkv.81">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5806" name="Constant_323526" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
+		<layer id="3351" name="__module.model.gpt_neox.layers.20.attention/prim::ListConstruct" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201115992" size="32" />
 			<output>
-				<port id="0" precision="I64" />
+				<port id="0" precision="I64">
+					<dim>4</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5807" name="Gather_323527" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="3352" name="__module.model.gpt_neox.layers.20.attention/aten::view/Reshape" type="Reshape" version="opset1">
+			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>7680</dim>
 				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
-				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="3" precision="FP32">
+				<port id="2" precision="FP32" names="2636,qkv.83">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5808" name="Constant_277928" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3353" name="Constant_13316" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5809" name="Constant_277931" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3354" name="Constant_13318" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5810" name="Constant_277934" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3355" name="Constant_13320" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5811" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3356" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Slice" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -88883,17 +50867,20 @@
 					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3761">
+				<port id="5" precision="FP32" names="2637">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -88901,15 +50888,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5812" name="Constant_13614" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="3357" name="Constant_13329" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64" names="3762">
+				<port id="0" precision="I64" names="2638">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5813" name="__module.model.gpt_neox.layers.23.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
+		<layer id="3358" name="__module.model.gpt_neox.layers.20.attention/aten::permute/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -88922,32 +50909,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3763">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5814" name="__module.model.gpt_neox.layers.23.attention/aten::cat/Concat_5" type="Concat" version="opset1">
-			<data axis="-2" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="present.23.value">
+				<port id="2" precision="FP32" names="2639,query.41">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -88955,41 +50917,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5815" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Slice_11" type="Const" version="opset1">
-			<data element_type="boolean" shape="1, 1, 2048, 2048" offset="201244276" size="4194304" />
+		<layer id="3359" name="Constant_13370" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="BOOL" names="3840">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5816" name="Constant_277938" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3360" name="Constant_13372" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>3</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5817" name="Constant_277937" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162298" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3361" name="Constant_13374" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5818" name="__module.model.gpt_neox.layers.23.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="3362" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Slice_1" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -88997,118 +50949,84 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-			</input>
-			<output>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="5819" name="__module.model.gpt_neox.layers.23.attention/aten::size/Constant_8" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="5820" name="__module.model.gpt_neox.layers.23.attention/aten::size/Gather_8" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="3835" />
+				<port id="5" precision="FP32" names="2646,q.41">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5821" name="__module.model.gpt_neox.layers.23.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="3363" name="__module.model.gpt_neox.layers.20.attention/aten::mul/Multiply" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="2652">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5822" name="__module.model.gpt_neox.layers.23.attention/aten::size/Constant_9" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3364" name="Constant_13528" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="5823" name="__module.model.gpt_neox.layers.23.attention/aten::size/Gather_9" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="3833" />
-			</output>
-		</layer>
-		<layer id="5824" name="__module.model.gpt_neox.layers.23.attention/aten::sub/Subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" names="3837,3838" />
-			</output>
-		</layer>
-		<layer id="5825" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Constant_4" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
 			</output>
 		</layer>
-		<layer id="5826" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="3365" name="Constant_13530" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5827" name="Constant_277936" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3366" name="Constant_13532" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5828" name="ScatterUpdate_277939" type="ScatterUpdate" version="opset3">
+		<layer id="3367" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Slice_2" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -89116,146 +51034,89 @@
 				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
 				<port id="4" precision="I64">
-					<dim>3</dim>
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="5829" name="Constant_277940" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="5" precision="FP32" names="2662,x2.81">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5830" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="3368" name="Constant_37453" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5831" name="ScatterUpdate_277941" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
 					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
 					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5832" name="Constant_277944" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438612" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
 			</output>
 		</layer>
-		<layer id="5833" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3369" name="__module.model.gpt_neox.layers.20.attention/aten::neg/Multiply" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="BOOL">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+				<port id="1" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>3</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="BOOL" names="3841">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="5834" name="Constant_278017" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="2663">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5835" name="Constant_278018" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3370" name="Constant_13425" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5836" name="Constant_278014" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3371" name="Constant_13469" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5837" name="Constant_278013" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3372" name="Constant_13471" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5838" name="ScatterUpdate_278019" type="ScatterUpdate" version="opset3">
+		<layer id="3373" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Slice_3" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -89263,183 +51124,128 @@
 				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
 				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5839" name="Constant_278022" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5840" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="BOOL" names="3842,3843">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="5" precision="FP32" names="2657">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5841" name="__module.model.gpt_neox.layers.23.attention/aten::where/Convert_1_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="5842" name="__module.model.gpt_neox.layers.23.attention/aten::where/Convert_1" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="5843" name="__module.model.gpt_neox.layers.23.attention/aten::full/Convert" type="Const" version="opset1">
-			<data element_type="f32" shape="" offset="205438636" size="4" />
-			<output>
-				<port id="0" precision="FP32" names="3830" />
-			</output>
-		</layer>
-		<layer id="5844" name="__module.model.gpt_neox.layers.23.attention/aten::where/Select" type="Select" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3374" name="__module.model.gpt_neox.layers.20.attention/aten::cat/Concat" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
-				<port id="1" precision="FP32" />
-				<port id="2" precision="FP32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="3844,causal_mask.93">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="2665">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5845" name="Add_290476" type="Add" version="opset1">
+		<layer id="3375" name="__module.model.gpt_neox.layers.20.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>1</dim>
 					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="2666">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5846" name="ShapeOf_290477" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="3376" name="__module.model.gpt_neox.layers.20.attention/aten::add/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="2667">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5847" name="Constant_254674" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3377" name="Constant_13380" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5848" name="__module.model.gpt_neox.layers.23.attention/aten::size/Constant_10" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3378" name="Constant_13382" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="5849" name="__module.model.gpt_neox.layers.23.attention/aten::size/Gather_10" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I32" />
-			</input>
+			</output>
+		</layer>
+		<layer id="3379" name="Constant_13384" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="3" precision="I64" names="3828">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5850" name="__module.model.gpt_neox.layers.23.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="3380" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Slice_4" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -89450,93 +51256,154 @@
 				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64" names="3845">
-					<dim>4</dim>
+				<port id="5" precision="FP32" names="2647">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5851" name="__module.model.gpt_neox.layers.23.attention/aten::expand/Abs" type="Abs" version="opset1">
+		<layer id="3381" name="__module.model.gpt_neox.layers.20.attention/aten::cat/Concat_1" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="2685">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5852" name="Maximum_290478" type="Maximum" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3382" name="Constant_53031" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
+			<output>
+				<port id="0" precision="FP32" />
+			</output>
+		</layer>
+		<layer id="3383" name="Constant_53024" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3384" name="Constant_53026" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3385" name="Constant_53028" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3386" name="Concat_53029" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
+				<port id="4" precision="I64">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5853" name="__module.model.gpt_neox.layers.23.attention/aten::add/Add_2" type="Broadcast" version="opset3">
+		<layer id="3387" name="Broadcast_53032" type="Broadcast" version="opset3">
 			<data mode="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
+				<port id="0" precision="FP32" />
 				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3847">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5854" name="__module.model.gpt_neox.layers.23.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
-			<data causal="false" />
+		<layer id="3388" name="ReadValue_50235" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.20.keypresent.20.key" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>-1</dim>
+					<dim>0</dim>
 					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
+			</input>
+			<output>
+				<port id="1" precision="FP32" names="past_key_values.20.key">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-				<port id="2" precision="FP32">
+			</output>
+		</layer>
+		<layer id="3389" name="Constant_47898" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64" />
+			</output>
+		</layer>
+		<layer id="3390" name="Gather_47899" type="Gather" version="opset8">
+			<data batch_dims="0" />
+			<input>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
+				<port id="1" precision="I32">
 					<dim>-1</dim>
 				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3848,3849,sdpa_result.47,tensor.93">
+				<port id="3" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -89544,28 +51411,53 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5855" name="Constant_14055" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="3391" name="Constant_13334" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
-				<port id="0" precision="I64" names="3850">
-					<dim>4</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3392" name="Constant_13336" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3393" name="Constant_13338" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5856" name="__module.model.gpt_neox.layers.23.attention/aten::permute/Transpose_3" type="Transpose" version="opset1">
+		<layer id="3394" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Slice_5" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3851">
+				<port id="5" precision="FP32" names="2640">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -89573,19 +51465,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5857" name="__module.model.gpt_neox.layers.23.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438650" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3395" name="Constant_13347" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="0" precision="I64" names="2641">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5858" name="__module.model.gpt_neox.layers.23.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
-			<data special_zero="true" />
+		<layer id="3396" name="__module.model.gpt_neox.layers.20.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -89594,986 +51482,960 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>3</dim>
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3856">
+				<port id="2" precision="FP32" names="2642,key.81">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5859" name="self.model.gpt_neox.layers.23.attention.dense.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 2560" offset="2017352914" size="6553600" />
+		<layer id="3397" name="Constant_13390" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5860" name="Convert_22354" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="3398" name="Constant_13392" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5861" name="self.model.gpt_neox.layers.23.attention.dense.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="2023906514" size="2560" />
+		<layer id="3399" name="Constant_13394" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5862" name="Convert_22357" type="Convert" version="opset1">
-			<data destination_type="f16" />
+		<layer id="3400" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Slice_6" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
+				<port id="5" precision="FP32" names="2648,k.41">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5863" name="self.model.gpt_neox.layers.23.attention.dense.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="3401" name="__module.model.gpt_neox.layers.20.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
 					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32" names="2668">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5864" name="self.model.gpt_neox.layers.23.attention.dense.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="2023909074" size="5120" />
+		<layer id="3402" name="Constant_13671" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5865" name="self.model.gpt_neox.layers.23.attention.dense.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+		<layer id="3403" name="Constant_13673" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="3404" name="Constant_13675" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5866" name="self.model.gpt_neox.layers.23.attention.dense.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="3405" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Slice_7" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="5" precision="FP32" names="2678,x2.83">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5867" name="__module.model.gpt_neox.layers.23.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
+		<layer id="3406" name="Constant_37454" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3407" name="__module.model.gpt_neox.layers.20.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>2560</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="2679">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5868" name="Constant_299308_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2023914194" size="5120" />
+		<layer id="3408" name="Constant_13568" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5869" name="Constant_299308" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+		<layer id="3409" name="Constant_13612" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="3410" name="Constant_13614" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5870" name="__module.model.gpt_neox.layers.23.attention.dense/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3411" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Slice_8" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="FP32">
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3859,3861,input.95">
+				<port id="5" precision="FP32" names="2673">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5871" name="__module.model.gpt_neox.layers.23/aten::add/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3412" name="__module.model.gpt_neox.layers.20.attention/aten::cat/Concat_2" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3879">
+				<port id="2" precision="FP32" names="2681">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5872" name="__module.model.gpt_neox.layers.23/aten::add/Add_1" type="Add" version="opset1">
+		<layer id="3413" name="__module.model.gpt_neox.layers.20.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3880,3882,hidden_states.49">
+				<port id="2" precision="FP32" names="2682">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5873" name="__module.model.gpt_neox.layers.24.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867670" size="4" />
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5874" name="__module.model.gpt_neox.layers.24.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
-			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
+		<layer id="3414" name="__module.model.gpt_neox.layers.20.attention/aten::add/Add_1" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="I32">
-					<dim>1</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="2683">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5875" name="Constant_299309_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2023919314" size="5120" />
+		<layer id="3415" name="Constant_13400" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5876" name="Constant_299309" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+		<layer id="3416" name="Constant_13402" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="3417" name="Constant_13404" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5877" name="__module.model.gpt_neox.layers.24.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3418" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Slice_9" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="1" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="5878" name="Constant_299310_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2023924434" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="2" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="5879" name="Constant_299310" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
+				<port id="3" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+				<port id="5" precision="FP32" names="2649">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5880" name="__module.model.gpt_neox.layers.24.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3419" name="__module.model.gpt_neox.layers.20.attention/aten::cat/Concat_3" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4017">
+				<port id="2" precision="FP32" names="2687">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5881" name="self.model.gpt_neox.layers.24.mlp.dense_h_to_4h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 2560" offset="2023929554" size="26214400" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5882" name="Convert_22404" type="Convert" version="opset1">
-			<data destination_type="f16" />
+		<layer id="3420" name="__module.model.gpt_neox.layers.20.attention/aten::cat/Concat_4" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32" names="present.20.key">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5883" name="self.model.gpt_neox.layers.24.mlp.dense_h_to_4h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 1" offset="2050143954" size="10240" />
+		<layer id="3421" name="Constant_53043" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
+				<port id="0" precision="FP32" />
 			</output>
 		</layer>
-		<layer id="5884" name="Convert_22407" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</input>
+		<layer id="3422" name="Constant_53036" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5885" name="self.model.gpt_neox.layers.24.mlp.dense_h_to_4h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</input>
+		<layer id="3423" name="Constant_53038" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="2" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5886" name="self.model.gpt_neox.layers.24.mlp.dense_h_to_4h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="10240, 1" offset="2050154194" size="20480" />
+		<layer id="3424" name="Constant_53040" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5887" name="self.model.gpt_neox.layers.24.mlp.dense_h_to_4h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3425" name="Concat_53041" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
+				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="2" precision="I64">
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="5888" name="self.model.gpt_neox.layers.24.mlp.dense_h_to_4h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="3" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="4" precision="I64">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5889" name="__module.model.gpt_neox.layers.24.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
+		<layer id="3426" name="Broadcast_53044" type="Broadcast" version="opset3">
+			<data mode="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
 				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5890" name="Constant_299311_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 10240" offset="2050174674" size="20480" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5891" name="Constant_299311" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5892" name="__module.model.gpt_neox.layers.24.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3427" name="ReadValue_50237" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.20.valuepresent.20.value" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4022">
+				<port id="1" precision="FP32" names="past_key_values.20.value">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5893" name="__module.model.gpt_neox.layers.24.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
-			<data approximation_mode="ERF" />
+		<layer id="3428" name="Constant_47901" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64" />
+			</output>
+		</layer>
+		<layer id="3429" name="Gather_47902" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I32">
 					<dim>-1</dim>
-					<dim>10240</dim>
 				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="4023">
+				<port id="3" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5894" name="self.model.gpt_neox.layers.24.mlp.dense_4h_to_h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 10240" offset="2050195154" size="26214400" />
+		<layer id="3430" name="Constant_13352" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5895" name="Convert_22414" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
+		<layer id="3431" name="Constant_13354" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5896" name="self.model.gpt_neox.layers.24.mlp.dense_4h_to_h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="2076409554" size="2560" />
+		<layer id="3432" name="Constant_13356" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5897" name="Convert_22417" type="Convert" version="opset1">
-			<data destination_type="f16" />
+		<layer id="3433" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Slice_10" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>240</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="5898" name="self.model.gpt_neox.layers.24.mlp.dense_4h_to_h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="3" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="5" precision="FP32" names="2643">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5899" name="self.model.gpt_neox.layers.24.mlp.dense_4h_to_h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="2076412114" size="5120" />
+		<layer id="3434" name="Constant_13365" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
+				<port id="0" precision="I64" names="2644">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5900" name="self.model.gpt_neox.layers.24.mlp.dense_4h_to_h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3435" name="__module.model.gpt_neox.layers.20.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="2" precision="FP32" names="2645">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5901" name="self.model.gpt_neox.layers.24.mlp.dense_4h_to_h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="3436" name="__module.model.gpt_neox.layers.20.attention/aten::cat/Concat_5" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="2" precision="FP32" names="present.20.value">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5902" name="__module.model.gpt_neox.layers.24.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
+		<layer id="3437" name="Constant_13747" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3438" name="ShapeOf_40132" type="ShapeOf" version="opset3">
+			<data output_type="i64" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5903" name="Constant_299312_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2076417234" size="5120" />
+		<layer id="3439" name="Constant_40133" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116168" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5904" name="Constant_299312" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="3440" name="Constant_40134" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64" />
+			</output>
+		</layer>
+		<layer id="3441" name="Gather_40135" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
+					<dim>4</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="3" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5905" name="__module.model.gpt_neox.layers.24.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1">
+		<layer id="3442" name="Add_40138" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4026,input.101">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
+				<port id="2" precision="I64" names="2692">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5906" name="Constant_299313_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2076422354" size="5120" />
+		<layer id="3443" name="Constant_13750" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5907" name="Constant_299313" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="3444" name="__module.model.gpt_neox.layers.20.attention/aten::slice/Slice_14" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
 					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="5" precision="FP32" names="2696">
+					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5908" name="__module.model.gpt_neox.layers.24.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3445" name="__module.model.gpt_neox.layers.20.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+			<data causal="false" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="3" precision="FP32">
+					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="4" precision="FP32" names="2697,attn_output.121">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5909" name="Constant_299314_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2076427474" size="5120" />
+		<layer id="3446" name="__module.model.gpt_neox.layers.20.attention/aten::transpose/Constant" type="Const" version="opset1">
+			<data element_type="i32" shape="4" offset="201116200" size="16" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I32">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5910" name="Constant_299314" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="3447" name="__module.model.gpt_neox.layers.20.attention/aten::transpose/Transpose" type="Transpose" version="opset1">
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I32">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32" names="2698">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5911" name="__module.model.gpt_neox.layers.24.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3448" name="Constant_41004" type="Const" version="opset1">
+			<data element_type="i64" shape="3" offset="201116216" size="24" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>3</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3449" name="__module.model.gpt_neox.layers.20.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
+			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+				<port id="1" precision="I64">
+					<dim>3</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3892">
+				<port id="2" precision="FP32" names="2701">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5912" name="self.model.gpt_neox.layers.24.attention.query_key_value.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 2560" offset="2076432594" size="19660800" />
+		<layer id="3450" name="self.model.gpt_neox.layers.20.attention.dense.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 2560" offset="1777513040" size="6553600" />
 			<output>
 				<port id="0" precision="U8">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5913" name="Convert_22384" type="Convert" version="opset1">
+		<layer id="3451" name="Convert_75396" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP16">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5914" name="self.model.gpt_neox.layers.24.attention.query_key_value.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 1" offset="2096093394" size="7680" />
+		<layer id="3452" name="self.model.gpt_neox.layers.20.attention.dense.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="1784066640" size="2560" />
 			<output>
 				<port id="0" precision="U8">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5915" name="Convert_22387" type="Convert" version="opset1">
+		<layer id="3453" name="Convert_75399" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP16">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5916" name="self.model.gpt_neox.layers.24.attention.query_key_value.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="3454" name="self.model.gpt_neox.layers.20.attention.dense.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP16">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
 				<port id="2" precision="FP16">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5917" name="self.model.gpt_neox.layers.24.attention.query_key_value.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="7680, 1" offset="2096101074" size="15360" />
+		<layer id="3455" name="self.model.gpt_neox.layers.20.attention.dense.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="1784069200" size="5120" />
 			<output>
 				<port id="0" precision="FP16">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5918" name="self.model.gpt_neox.layers.24.attention.query_key_value.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="3456" name="self.model.gpt_neox.layers.20.attention.dense.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP16">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
 				<port id="2" precision="FP16">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5919" name="self.model.gpt_neox.layers.24.attention.query_key_value.weight" type="Convert" version="opset1">
+		<layer id="3457" name="self.model.gpt_neox.layers.20.attention.dense.weight/fq_weights_1/convert" type="Convert" version="opset1">
 			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
 			<input>
 				<port id="0" precision="FP16">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP32">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5920" name="__module.model.gpt_neox.layers.24.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="3458" name="__module.model.gpt_neox.layers.20.attention.dense/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -90582,7 +52444,7 @@
 					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
@@ -90590,826 +52452,838 @@
 				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5921" name="Constant_299315_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 7680" offset="2096116434" size="15360" />
+		<layer id="3459" name="Constant_37455" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1784074320" size="10240" />
 			<output>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5922" name="Constant_299315" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="3460" name="__module.model.gpt_neox.layers.20.attention.dense/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
+				<port id="2" precision="FP32" names="2704,2706,input.81">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5923" name="__module.model.gpt_neox.layers.24.attention.query_key_value/aten::linear/Add" type="Add" version="opset1">
+		<layer id="3461" name="__module.model.gpt_neox.layers.20/aten::add/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3899,qkv.97">
+				<port id="2" precision="FP32" names="2724">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5924" name="__module.model.gpt_neox.layers.24.attention/prim::ListConstruct/Concat" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080154" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5925" name="__module.model.gpt_neox.layers.24.attention/aten::view/Reshape" type="Reshape" version="opset1">
-			<data special_zero="true" />
+		<layer id="3462" name="__module.model.gpt_neox.layers.20/aten::add/Add_1" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3903,qkv.99">
+				<port id="2" precision="FP32" names="2725,2727,hidden_states.87">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5926" name="Constant_278029" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5927" name="Constant_278032" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5928" name="Constant_278035" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3463" name="__module.model.gpt_neox.layers.21.input_layernorm/aten::layer_norm/Multiply" type="Const" version="opset1">
+			<data element_type="i32" shape="1" offset="128867668" size="4" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="I32">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5929" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3464" name="__module.model.gpt_neox.layers.21.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
+			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="I32">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3904">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5930" name="Constant_14167" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="3465" name="Constant_37456" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1784084560" size="10240" />
 			<output>
-				<port id="0" precision="I64" names="3905">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5931" name="__module.model.gpt_neox.layers.24.attention/aten::permute/Transpose" type="Transpose" version="opset1">
+		<layer id="3466" name="__module.model.gpt_neox.layers.21.post_attention_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3906,query.145">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5932" name="Constant_278041" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3467" name="Constant_37457" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1784094800" size="10240" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5933" name="Constant_278044" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3468" name="__module.model.gpt_neox.layers.21.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="2825">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5934" name="Constant_278047" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3469" name="self.model.gpt_neox.layers.21.mlp.dense_h_to_4h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 2560" offset="1784105040" size="26214400" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5935" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3470" name="Convert_74604" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
-			<output>
-				<port id="4" precision="FP32" names="3913,q.49">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+			<output>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5936" name="self.model.gpt_neox.layers.24.attention.rotary_emb.cos_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201080346" size="81920" />
+		<layer id="3471" name="self.model.gpt_neox.layers.21.mlp.dense_h_to_4h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 1" offset="1810319440" size="10240" />
 			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.24.attention.rotary_emb.cos_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5937" name="self.model.gpt_neox.layers.24.attention.rotary_emb.cos_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="3472" name="Convert_74607" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5938" name="__module.model.gpt_neox.layers.24.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
+		<layer id="3473" name="self.model.gpt_neox.layers.21.mlp.dense_h_to_4h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5939" name="Constant_278053" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3474" name="self.model.gpt_neox.layers.21.mlp.dense_h_to_4h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="10240, 1" offset="1810329680" size="20480" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5940" name="Constant_278056" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3475" name="self.model.gpt_neox.layers.21.mlp.dense_h_to_4h.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5941" name="Constant_278059" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
+		<layer id="3476" name="__module.model.gpt_neox.layers.21.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<rt_info>
-				<attribute name="precise" version="0" />
+				<attribute name="decompression" version="0" />
 			</rt_info>
+			<input>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5942" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3477" name="__module.model.gpt_neox.layers.21.mlp.dense_h_to_4h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3907">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5943" name="Constant_14181" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="3478" name="self.model.gpt_neox.layers.21.mlp.dense_h_to_4h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="10240" offset="1810350160" size="20480" />
 			<output>
-				<port id="0" precision="I64" names="3908">
-					<dim>4</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.21.mlp.dense_h_to_4h.bias">
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3479" name="__module.model.gpt_neox.layers.21.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<rt_info>
+				<attribute name="decompression" version="0" />
+			</rt_info>
+			<input>
+				<port id="0" precision="BF16">
+					<dim>10240</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5944" name="__module.model.gpt_neox.layers.24.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
+		<layer id="3480" name="__module.model.gpt_neox.layers.21.mlp.dense_h_to_4h/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3909,key.145">
+				<port id="2" precision="FP32" names="2830">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5945" name="__module.model.gpt_neox.layers.24.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="3481" name="__module.model.gpt_neox.layers.21.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
+			<data approximation_mode="ERF" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32" names="2831">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5946" name="__module.model.gpt_neox.layers.24.attention/aten::size/Constant_2" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3482" name="self.model.gpt_neox.layers.21.mlp.dense_4h_to_h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 10240" offset="1810370640" size="26214400" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5947" name="__module.model.gpt_neox.layers.24.attention/aten::size/Gather_2" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="3483" name="Convert_74614" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="3917" />
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5948" name="Constant_338053_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
+		<layer id="3484" name="self.model.gpt_neox.layers.21.mlp.dense_4h_to_h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="1836585040" size="2560" />
 			<output>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5949" name="Constant_338053" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="3485" name="Convert_74617" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="5950" name="Constant_338046" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
-			<output>
-				<port id="0" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5951" name="Constant_338048" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64">
+		<layer id="3486" name="self.model.gpt_neox.layers.21.mlp.dense_4h_to_h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5952" name="Constant_338050" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
+		<layer id="3487" name="self.model.gpt_neox.layers.21.mlp.dense_4h_to_h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="1836587600" size="5120" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5953" name="Concat_338051" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="3488" name="self.model.gpt_neox.layers.21.mlp.dense_4h_to_h.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="3" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5954" name="Broadcast_338054" type="Broadcast" version="opset3">
-			<data mode="numpy" />
+		<layer id="3489" name="__module.model.gpt_neox.layers.21.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<rt_info>
+				<attribute name="decompression" version="0" />
+			</rt_info>
 			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5955" name="ReadValue_330493" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.24.keypresent.24.key" variable_type="f32" variable_shape="?,32,?,80" />
+		<layer id="3490" name="__module.model.gpt_neox.layers.21.mlp.dense_4h_to_h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="past_key_values.24.key">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5956" name="Constant_323529" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
+		<layer id="3491" name="self.model.gpt_neox.layers.21.mlp.dense_4h_to_h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="2560" offset="1836592720" size="5120" />
 			<output>
-				<port id="0" precision="I64" />
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.21.mlp.dense_4h_to_h.bias">
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5957" name="Gather_323530" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="3492" name="__module.model.gpt_neox.layers.21.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<rt_info>
+				<attribute name="decompression" version="0" />
+			</rt_info>
+			<input>
+				<port id="0" precision="BF16">
+					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3493" name="__module.model.gpt_neox.layers.21.mlp.dense_4h_to_h/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
 				</port>
-				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="3" precision="FP32">
+				<port id="2" precision="FP32" names="2834,input.87">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5958" name="__module.model.gpt_neox.layers.24.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="3494" name="Constant_37458" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1836597840" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3495" name="__module.model.gpt_neox.layers.21.input_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5959" name="__module.model.gpt_neox.layers.24.attention/aten::size/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3496" name="Constant_37459" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1836608080" size="10240" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="5960" name="__module.model.gpt_neox.layers.24.attention/aten::size/Gather_3" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="3919" />
 			</output>
 		</layer>
-		<layer id="5961" name="__module.model.gpt_neox.layers.24.attention/aten::add_/Add" type="Add" version="opset1">
+		<layer id="3497" name="__module.model.gpt_neox.layers.21.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64" names="3918,3922,3923,seq_len.97" />
+				<port id="2" precision="FP32" names="2737,hidden_states.85">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5962" name="__module.model.gpt_neox.layers.24.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3498" name="self.model.gpt_neox.layers.21.attention.query_key_value.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 2560" offset="1836618320" size="19660800" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5963" name="__module.model.gpt_neox.layers.24.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1">
+		<layer id="3499" name="Convert_75055" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5964" name="__module.model.gpt_neox.layers.24.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3500" name="self.model.gpt_neox.layers.21.attention.query_key_value.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 1" offset="1856279120" size="7680" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="U8">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5965" name="__module.model.gpt_neox.layers.24.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3501" name="Convert_75058" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
+				<port id="0" precision="U8">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
+			</input>
+			<output>
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I64">
+			</output>
+		</layer>
+		<layer id="3502" name="self.model.gpt_neox.layers.21.attention.query_key_value.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3926,3929,cos.97">
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="2" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5966" name="Constant_38341" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
+		<layer id="3503" name="self.model.gpt_neox.layers.21.attention.query_key_value.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="7680, 1" offset="1856286800" size="15360" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5967" name="__module.model.gpt_neox.layers.24.attention/aten::index/Gather" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="3504" name="self.model.gpt_neox.layers.21.attention.query_key_value.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
+					<dim>1</dim>
 				</port>
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="3932">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="2" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5968" name="__module.model.gpt_neox.layers.24.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1">
+		<layer id="3505" name="self.model.gpt_neox.layers.21.attention.query_key_value.weight/fq_weights_1/convert" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3933,cos.99">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="1" precision="FP32">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5969" name="__module.model.gpt_neox.layers.24.attention/aten::mul/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3506" name="__module.model.gpt_neox.layers.21.attention.query_key_value/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3937">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5970" name="Constant_278064" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5971" name="Constant_278063" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3507" name="Constant_37460" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 7680" offset="1856302160" size="30720" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="FP32">
+					<dim>1</dim>
 					<dim>1</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5972" name="__module.model.gpt_neox.layers.24.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="3508" name="__module.model.gpt_neox.layers.21.attention.query_key_value/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>7680</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>7680</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="2744,qkv.85">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5973" name="__module.model.gpt_neox.layers.24.attention/aten::size/Constant_5" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3509" name="__module.model.gpt_neox.layers.21.attention/prim::ListConstruct" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201115992" size="32" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="5974" name="__module.model.gpt_neox.layers.24.attention/aten::size/Gather_5" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
 				<port id="0" precision="I64">
 					<dim>4</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="3938,3943" />
-			</output>
-		</layer>
-		<layer id="5975" name="__module.model.gpt_neox.layers.24.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" />
 			</output>
 		</layer>
-		<layer id="5976" name="__module.model.gpt_neox.layers.24.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1">
+		<layer id="3510" name="__module.model.gpt_neox.layers.21.attention/aten::view/Reshape" type="Reshape" version="opset1">
+			<data special_zero="true" />
 			<input>
-				<port id="0" precision="I64" />
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>7680</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>4</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64" names="3940,3941,3945,3946" />
+				<port id="2" precision="FP32" names="2748,qkv.87">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>240</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5977" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Constant_1" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3511" name="Constant_13946" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="5978" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="3512" name="Constant_13948" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5979" name="Constant_278062" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3513" name="Constant_13950" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5980" name="ScatterUpdate_278065" type="ScatterUpdate" version="opset3">
+		<layer id="3514" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Slice" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -91417,148 +53291,155 @@
 				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
 				<port id="4" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="5981" name="Constant_278068" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="5" precision="FP32" names="2749">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5982" name="Constant_278071" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3515" name="Constant_13959" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="I64" names="2750">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5983" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3516" name="__module.model.gpt_neox.layers.21.attention/aten::permute/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3947,x2.97">
+				<port id="2" precision="FP32" names="2751,query.43">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5984" name="Constant_299316_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
+		<layer id="3517" name="Constant_14000" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3518" name="Constant_14002" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3519" name="Constant_14004" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5985" name="Constant_299316" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="3520" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Slice_1" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="5" precision="FP32" names="2758,q.43">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5986" name="__module.model.gpt_neox.layers.24.attention/aten::neg/Multiply" type="Multiply" version="opset1">
+		<layer id="3521" name="__module.model.gpt_neox.layers.21.attention/aten::mul/Multiply" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
+					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3948">
+				<port id="2" precision="FP32" names="2764">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5987" name="Constant_278116" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3522" name="Constant_14158" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5988" name="Constant_278121" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3523" name="Constant_14160" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3524" name="Constant_14162" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5989" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3525" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Slice_2" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -91567,17 +53448,20 @@
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3942">
+				<port id="5" precision="FP32" names="2774,x2.85">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -91585,8 +53469,19 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="5990" name="__module.model.gpt_neox.layers.24.attention/aten::cat/Concat" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="3526" name="Constant_37461" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3527" name="__module.model.gpt_neox.layers.21.attention/aten::neg/Multiply" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -91595,75 +53490,51 @@
 					<dim>10</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3950">
+				<port id="2" precision="FP32" names="2775">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="5991" name="self.model.gpt_neox.layers.24.attention.rotary_emb.sin_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201162356" size="81920" />
-			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.24.attention.rotary_emb.sin_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5992" name="self.model.gpt_neox.layers.24.attention.rotary_emb.sin_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</input>
+		<layer id="3528" name="Constant_14055" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5993" name="__module.model.gpt_neox.layers.24.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3529" name="Constant_14099" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5994" name="__module.model.gpt_neox.layers.24.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3530" name="Constant_14101" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5995" name="__module.model.gpt_neox.layers.24.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3531" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Slice_3" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>2048</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
@@ -91675,60 +53546,45 @@
 				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3927,3930,sin.97">
+				<port id="5" precision="FP32" names="2769">
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5996" name="Constant_38344" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="5997" name="__module.model.gpt_neox.layers.24.attention/aten::index/Gather_1" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="3532" name="__module.model.gpt_neox.layers.21.attention/aten::cat/Concat" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="3935">
-					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>10</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="5998" name="__module.model.gpt_neox.layers.24.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
+				<port id="1" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>10</dim>
 				</port>
-				<port id="1" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3936,sin.99">
+				<port id="2" precision="FP32" names="2777">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="5999" name="__module.model.gpt_neox.layers.24.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
+		<layer id="3533" name="__module.model.gpt_neox.layers.21.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -91745,7 +53601,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3951">
+				<port id="2" precision="FP32" names="2778">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -91753,7 +53609,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6000" name="__module.model.gpt_neox.layers.24.attention/aten::add/Add" type="Add" version="opset1">
+		<layer id="3534" name="__module.model.gpt_neox.layers.21.attention/aten::add/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -91770,7 +53626,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3952">
+				<port id="2" precision="FP32" names="2779">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -91778,41 +53634,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6001" name="Constant_278167" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3535" name="Constant_14010" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6002" name="Constant_278170" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3536" name="Constant_14012" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6003" name="Constant_278173" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3537" name="Constant_14014" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6004" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3538" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Slice_4" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -91821,17 +53667,20 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3914">
+				<port id="5" precision="FP32" names="2759">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -91839,7 +53688,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6005" name="__module.model.gpt_neox.layers.24.attention/aten::cat/Concat_1" type="Concat" version="opset1">
+		<layer id="3539" name="__module.model.gpt_neox.layers.21.attention/aten::cat/Concat_1" type="Concat" version="opset1">
 			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
@@ -91856,7 +53705,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3970,3980,query.147,query.149">
+				<port id="2" precision="FP32" names="2797">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -91864,205 +53713,237 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6006" name="Constant_278179" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3540" name="Constant_53055" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
+			<output>
+				<port id="0" precision="FP32" />
+			</output>
+		</layer>
+		<layer id="3541" name="Constant_53048" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6007" name="Constant_278182" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3542" name="Constant_53050" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6008" name="Constant_278185" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3543" name="Constant_53052" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6009" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3544" name="Concat_53053" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="4" precision="I64">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3545" name="Broadcast_53056" type="Broadcast" version="opset3">
+			<data mode="numpy" />
+			<input>
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3915,k.49">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6010" name="__module.model.gpt_neox.layers.24.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3546" name="ReadValue_50239" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.21.keypresent.21.key" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
 				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP32" names="past_key_values.21.key">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
+			</output>
+		</layer>
+		<layer id="3547" name="Constant_47904" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64" />
+			</output>
+		</layer>
+		<layer id="3548" name="Gather_47905" type="Gather" version="opset8">
+			<data batch_dims="0" />
+			<input>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I32">
 					<dim>-1</dim>
-					<dim>20</dim>
 				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3953">
+				<port id="3" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6011" name="Constant_278189" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3549" name="Constant_13964" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6012" name="Constant_278188" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3550" name="Constant_13966" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6013" name="__module.model.gpt_neox.layers.24.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="3551" name="Constant_13968" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3552" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Slice_5" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
-			</input>
-			<output>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="6014" name="__module.model.gpt_neox.layers.24.attention/aten::size/Constant_7" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="6015" name="__module.model.gpt_neox.layers.24.attention/aten::size/Gather_7" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="3954,3959" />
+				<port id="5" precision="FP32" names="2752">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6016" name="__module.model.gpt_neox.layers.24.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
+		<layer id="3553" name="Constant_13977" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="2" precision="I64" />
+				<port id="0" precision="I64" names="2753">
+					<dim>4</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6017" name="__module.model.gpt_neox.layers.24.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1">
+		<layer id="3554" name="__module.model.gpt_neox.layers.21.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
 			<input>
-				<port id="0" precision="I64" />
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>4</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64" names="3956,3957,3961,3962" />
+				<port id="2" precision="FP32" names="2754,key.85">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6018" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3555" name="Constant_14020" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6019" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="3556" name="Constant_14022" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6020" name="Constant_278187" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3557" name="Constant_14024" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6021" name="ScatterUpdate_278190" type="ScatterUpdate" version="opset3">
+		<layer id="3558" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Slice_6" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -92070,40 +53951,72 @@
 				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="5" precision="FP32" names="2760,k.43">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6022" name="Constant_278193" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3559" name="__module.model.gpt_neox.layers.21.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="2780">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3560" name="Constant_14301" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6023" name="Constant_278196" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3561" name="Constant_14303" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3562" name="Constant_14305" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6024" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3563" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Slice_7" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -92112,17 +54025,20 @@
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3963,x2.99">
+				<port id="5" precision="FP32" names="2790,x2.87">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -92130,32 +54046,10 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6025" name="Constant_299317_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6026" name="Constant_299317" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
+		<layer id="3564" name="Constant_37462" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>1</dim>
@@ -92163,7 +54057,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6027" name="__module.model.gpt_neox.layers.24.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
+		<layer id="3565" name="__module.model.gpt_neox.layers.21.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -92180,7 +54074,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3964">
+				<port id="2" precision="FP32" names="2791">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -92188,30 +54082,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6028" name="Constant_278241" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3566" name="Constant_14198" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6029" name="Constant_278246" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3567" name="Constant_14242" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3568" name="Constant_14244" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6030" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3569" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Slice_8" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -92220,17 +54115,20 @@
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3958">
+				<port id="5" precision="FP32" names="2785">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -92238,7 +54136,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6031" name="__module.model.gpt_neox.layers.24.attention/aten::cat/Concat_2" type="Concat" version="opset1">
+		<layer id="3570" name="__module.model.gpt_neox.layers.21.attention/aten::cat/Concat_2" type="Concat" version="opset1">
 			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
@@ -92255,7 +54153,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3966">
+				<port id="2" precision="FP32" names="2793">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -92263,7 +54161,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6032" name="__module.model.gpt_neox.layers.24.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
+		<layer id="3571" name="__module.model.gpt_neox.layers.21.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -92280,7 +54178,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3967">
+				<port id="2" precision="FP32" names="2794">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -92288,7 +54186,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6033" name="__module.model.gpt_neox.layers.24.attention/aten::add/Add_1" type="Add" version="opset1">
+		<layer id="3572" name="__module.model.gpt_neox.layers.21.attention/aten::add/Add_1" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -92305,7 +54203,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3968">
+				<port id="2" precision="FP32" names="2795">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -92313,41 +54211,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6034" name="Constant_278291" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3573" name="Constant_14030" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6035" name="Constant_278294" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3574" name="Constant_14032" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6036" name="Constant_278297" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3575" name="Constant_14034" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6037" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3576" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Slice_9" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -92356,17 +54244,20 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3916">
+				<port id="5" precision="FP32" names="2761">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -92374,7 +54265,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6038" name="__module.model.gpt_neox.layers.24.attention/aten::cat/Concat_3" type="Concat" version="opset1">
+		<layer id="3577" name="__module.model.gpt_neox.layers.21.attention/aten::cat/Concat_3" type="Concat" version="opset1">
 			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
@@ -92391,7 +54282,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3972">
+				<port id="2" precision="FP32" names="2799">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -92399,7 +54290,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6039" name="__module.model.gpt_neox.layers.24.attention/aten::to/Convert_1" type="Concat" version="opset1">
+		<layer id="3578" name="__module.model.gpt_neox.layers.21.attention/aten::cat/Concat_4" type="Concat" version="opset1">
 			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
@@ -92416,7 +54307,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="present.24.key">
+				<port id="2" precision="FP32" names="present.21.key">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -92424,49 +54315,37 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6040" name="Constant_338066_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="6041" name="Constant_338066" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
+		<layer id="3579" name="Constant_53067" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="1" precision="FP32" />
+				<port id="0" precision="FP32" />
 			</output>
 		</layer>
-		<layer id="6042" name="Constant_338059" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
+		<layer id="3580" name="Constant_53060" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6043" name="Constant_338061" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
+		<layer id="3581" name="Constant_53062" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6044" name="Constant_338063" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
+		<layer id="3582" name="Constant_53064" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6045" name="Concat_338064" type="Concat" version="opset1">
+		<layer id="3583" name="Concat_53065" type="Concat" version="opset1">
 			<data axis="0" />
 			<input>
 				<port id="0" precision="I64">
@@ -92488,7 +54367,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6046" name="Broadcast_338067" type="Broadcast" version="opset3">
+		<layer id="3584" name="Broadcast_53068" type="Broadcast" version="opset3">
 			<data mode="numpy" />
 			<input>
 				<port id="0" precision="FP32" />
@@ -92505,8 +54384,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6047" name="ReadValue_330495" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.24.valuepresent.24.value" variable_type="f32" variable_shape="?,32,?,80" />
+		<layer id="3585" name="ReadValue_50241" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.21.valuepresent.21.value" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -92516,7 +54395,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="past_key_values.24.value">
+				<port id="1" precision="FP32" names="past_key_values.21.value">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -92524,13 +54403,13 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6048" name="Constant_323532" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
+		<layer id="3586" name="Constant_47907" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="6049" name="Gather_323533" type="Gather" version="opset8">
+		<layer id="3587" name="Gather_47908" type="Gather" version="opset8">
 			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="FP32">
@@ -92553,41 +54432,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6050" name="Constant_278303" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3588" name="Constant_13982" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6051" name="Constant_278306" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3589" name="Constant_13984" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6052" name="Constant_278309" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3590" name="Constant_13986" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6053" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3591" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Slice_10" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -92596,17 +54465,20 @@
 					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3910">
+				<port id="5" precision="FP32" names="2755">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -92614,15 +54486,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6054" name="Constant_14195" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="3592" name="Constant_13995" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64" names="3911">
+				<port id="0" precision="I64" names="2756">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6055" name="__module.model.gpt_neox.layers.24.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
+		<layer id="3593" name="__module.model.gpt_neox.layers.21.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -92635,7 +54507,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3912">
+				<port id="2" precision="FP32" names="2757">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -92643,7 +54515,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6056" name="__module.model.gpt_neox.layers.24.attention/aten::cat/Concat_5" type="Concat" version="opset1">
+		<layer id="3594" name="__module.model.gpt_neox.layers.21.attention/aten::cat/Concat_5" type="Concat" version="opset1">
 			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
@@ -92660,7 +54532,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="present.24.value">
+				<port id="2" precision="FP32" names="present.21.value">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -92668,78 +54540,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6057" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Slice_11" type="Const" version="opset1">
-			<data element_type="boolean" shape="1, 1, 2048, 2048" offset="201244276" size="4194304" />
-			<output>
-				<port id="0" precision="BOOL" names="3989">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6058" name="Constant_278313" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6059" name="Constant_278312" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162298" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3595" name="Constant_14377" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6060" name="__module.model.gpt_neox.layers.24.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6061" name="__module.model.gpt_neox.layers.24.attention/aten::size/Constant_8" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="6062" name="__module.model.gpt_neox.layers.24.attention/aten::size/Gather_8" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="3984" />
-			</output>
-		</layer>
-		<layer id="6063" name="__module.model.gpt_neox.layers.24.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3">
+		<layer id="3596" name="ShapeOf_40203" type="ShapeOf" version="opset3">
 			<data output_type="i64" />
 			<input>
 				<port id="0" precision="FP32">
@@ -92755,348 +54564,62 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6064" name="__module.model.gpt_neox.layers.24.attention/aten::size/Constant_9" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="6065" name="__module.model.gpt_neox.layers.24.attention/aten::size/Gather_9" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="3982" />
-			</output>
-		</layer>
-		<layer id="6066" name="__module.model.gpt_neox.layers.24.attention/aten::sub/Subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" names="3986,3987" />
-			</output>
-		</layer>
-		<layer id="6067" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Constant_4" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="6068" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6069" name="Constant_278311" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3597" name="Constant_40204" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116168" size="8" />
 			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6070" name="ScatterUpdate_278314" type="ScatterUpdate" version="opset3">
-			<input>
 				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
-				</port>
 			</output>
 		</layer>
-		<layer id="6071" name="Constant_278315" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3598" name="Constant_40205" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6072" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1">
-			<input>
 				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
 			</output>
 		</layer>
-		<layer id="6073" name="ScatterUpdate_278316" type="ScatterUpdate" version="opset3">
+		<layer id="3599" name="Gather_40206" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="I64">
-					<dim>3</dim>
+					<dim>4</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6074" name="Constant_278319" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438612" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6075" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>3</dim>
-				</port>
 				<port id="3" precision="I64">
-					<dim>3</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="BOOL" names="3990">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6076" name="Constant_278392" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6077" name="Constant_278393" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6078" name="Constant_278389" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6079" name="Constant_278388" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6080" name="ScatterUpdate_278394" type="ScatterUpdate" version="opset3">
+		<layer id="3600" name="Add_40209" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6081" name="Constant_278397" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6082" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="BOOL" names="3991,3992">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6083" name="__module.model.gpt_neox.layers.24.attention/aten::where/Convert_1_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="6084" name="__module.model.gpt_neox.layers.24.attention/aten::where/Convert_1" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="6085" name="__module.model.gpt_neox.layers.24.attention/aten::full/Convert" type="Const" version="opset1">
-			<data element_type="f32" shape="" offset="205438636" size="4" />
-			<output>
-				<port id="0" precision="FP32" names="3979" />
-			</output>
-		</layer>
-		<layer id="6086" name="__module.model.gpt_neox.layers.24.attention/aten::where/Select" type="Select" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
 				</port>
-				<port id="1" precision="FP32" />
-				<port id="2" precision="FP32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="3993,causal_mask.97">
+				<port id="2" precision="I64" names="2804">
 					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6087" name="Add_290482" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-				</port>
-			</input>
+		<layer id="3601" name="Constant_14380" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6088" name="ShapeOf_290483" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="3602" name="__module.model.gpt_neox.layers.21.attention/aten::slice/Slice_14" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -93104,56 +54627,6 @@
 					<dim>-1</dim>
 					<dim>-1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6089" name="Constant_254689" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6090" name="__module.model.gpt_neox.layers.24.attention/aten::size/Constant_10" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="6091" name="__module.model.gpt_neox.layers.24.attention/aten::size/Gather_10" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="3977">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6092" name="__module.model.gpt_neox.layers.24.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1">
-			<data axis="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
@@ -93163,64 +54636,20 @@
 				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64" names="3994">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6093" name="__module.model.gpt_neox.layers.24.attention/aten::expand/Abs" type="Abs" version="opset1">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6094" name="Maximum_290484" type="Maximum" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6095" name="__module.model.gpt_neox.layers.24.attention/aten::add/Add_2" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="3996">
-					<dim>-1</dim>
+				<port id="5" precision="FP32" names="2808">
 					<dim>-1</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
 					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6096" name="__module.model.gpt_neox.layers.24.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+		<layer id="3603" name="__module.model.gpt_neox.layers.21.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
 			<data causal="false" />
 			<input>
 				<port id="0" precision="FP32">
@@ -93243,13 +54672,13 @@
 				</port>
 				<port id="3" precision="FP32">
 					<dim>-1</dim>
-					<dim>-1</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
 					<dim>-1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="3997,3998,sdpa_result.49,tensor.97">
+				<port id="4" precision="FP32" names="2809,attn_output.127">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -93257,15 +54686,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6097" name="Constant_14636" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="3604" name="__module.model.gpt_neox.layers.21.attention/aten::transpose/Constant" type="Const" version="opset1">
+			<data element_type="i32" shape="4" offset="201116200" size="16" />
 			<output>
-				<port id="0" precision="I64" names="3999">
+				<port id="0" precision="I32">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6098" name="__module.model.gpt_neox.layers.24.attention/aten::permute/Transpose_3" type="Transpose" version="opset1">
+		<layer id="3605" name="__module.model.gpt_neox.layers.21.attention/aten::transpose/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -93273,12 +54702,12 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="I32">
 					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4000">
+				<port id="2" precision="FP32" names="2810">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -93286,18 +54715,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6099" name="__module.model.gpt_neox.layers.24.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438650" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3606" name="Constant_41005" type="Const" version="opset1">
+			<data element_type="i64" shape="3" offset="201116216" size="24" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>3</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6100" name="__module.model.gpt_neox.layers.24.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
+		<layer id="3607" name="__module.model.gpt_neox.layers.21.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
 			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -93311,15 +54737,15 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4005">
+				<port id="2" precision="FP32" names="2813">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6101" name="self.model.gpt_neox.layers.24.attention.dense.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 2560" offset="2096131794" size="6553600" />
+		<layer id="3608" name="self.model.gpt_neox.layers.21.attention.dense.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 2560" offset="1856332880" size="6553600" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>2560</dim>
@@ -93327,7 +54753,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6102" name="Convert_22394" type="Convert" version="opset1">
+		<layer id="3609" name="Convert_75407" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -93342,8 +54768,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6103" name="self.model.gpt_neox.layers.24.attention.dense.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="2102685394" size="2560" />
+		<layer id="3610" name="self.model.gpt_neox.layers.21.attention.dense.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="1862886480" size="2560" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>2560</dim>
@@ -93351,7 +54777,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6104" name="Convert_22397" type="Convert" version="opset1">
+		<layer id="3611" name="Convert_75410" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -93366,7 +54792,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6105" name="self.model.gpt_neox.layers.24.attention.dense.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="3612" name="self.model.gpt_neox.layers.21.attention.dense.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -93385,8 +54811,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6106" name="self.model.gpt_neox.layers.24.attention.dense.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="2102687954" size="5120" />
+		<layer id="3613" name="self.model.gpt_neox.layers.21.attention.dense.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="1862889040" size="5120" />
 			<output>
 				<port id="0" precision="FP16">
 					<dim>2560</dim>
@@ -93394,7 +54820,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6107" name="self.model.gpt_neox.layers.24.attention.dense.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="3614" name="self.model.gpt_neox.layers.21.attention.dense.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -93413,11 +54839,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6108" name="self.model.gpt_neox.layers.24.attention.dense.weight" type="Convert" version="opset1">
+		<layer id="3615" name="self.model.gpt_neox.layers.21.attention.dense.weight/fq_weights_1/convert" type="Convert" version="opset1">
 			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
 			<input>
 				<port id="0" precision="FP16">
 					<dim>2560</dim>
@@ -93431,7 +54854,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6109" name="__module.model.gpt_neox.layers.24.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="3616" name="__module.model.gpt_neox.layers.21.attention.dense/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -93452,37 +54875,17 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6110" name="Constant_299318_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2102693074" size="5120" />
+		<layer id="3617" name="Constant_37463" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1862894160" size="10240" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6111" name="Constant_299318" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6112" name="__module.model.gpt_neox.layers.24.attention.dense/aten::linear/Add" type="Add" version="opset1">
+		<layer id="3618" name="__module.model.gpt_neox.layers.21.attention.dense/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -93497,14 +54900,14 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4008,4010,input.99">
+				<port id="2" precision="FP32" names="2816,2818,input.85">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6113" name="__module.model.gpt_neox.layers.24/aten::add/Add" type="Add" version="opset1">
+		<layer id="3619" name="__module.model.gpt_neox.layers.21/aten::add/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -93519,14 +54922,14 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4028">
+				<port id="2" precision="FP32" names="2836">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6114" name="__module.model.gpt_neox.layers.24/aten::add/Add_1" type="Add" version="opset1">
+		<layer id="3620" name="__module.model.gpt_neox.layers.21/aten::add/Add_1" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -93541,22 +54944,22 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4029,4031,hidden_states.51">
+				<port id="2" precision="FP32" names="2837,2839,hidden_states.91">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6115" name="__module.model.gpt_neox.layers.25.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867670" size="4" />
+		<layer id="3621" name="__module.model.gpt_neox.layers.22.input_layernorm/aten::layer_norm/Multiply" type="Const" version="opset1">
+			<data element_type="i32" shape="1" offset="128867668" size="4" />
 			<output>
 				<port id="0" precision="I32">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6116" name="__module.model.gpt_neox.layers.25.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
+		<layer id="3622" name="__module.model.gpt_neox.layers.22.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
 			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
 			<input>
 				<port id="0" precision="FP32">
@@ -93576,37 +54979,17 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6117" name="Constant_299319_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2102698194" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6118" name="Constant_299319" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="3623" name="Constant_37464" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1862904400" size="10240" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6119" name="__module.model.gpt_neox.layers.25.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
+		<layer id="3624" name="__module.model.gpt_neox.layers.22.post_attention_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -93628,37 +55011,17 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6120" name="Constant_299320_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2102703314" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6121" name="Constant_299320" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="3625" name="Constant_37465" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1862914640" size="10240" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6122" name="__module.model.gpt_neox.layers.25.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+		<layer id="3626" name="__module.model.gpt_neox.layers.22.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -93673,15 +55036,15 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4166">
+				<port id="2" precision="FP32" names="2937">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6123" name="self.model.gpt_neox.layers.25.mlp.dense_h_to_4h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 2560" offset="2102708434" size="26214400" />
+		<layer id="3627" name="self.model.gpt_neox.layers.22.mlp.dense_h_to_4h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 2560" offset="1862924880" size="26214400" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>10240</dim>
@@ -93689,7 +55052,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6124" name="Convert_22444" type="Convert" version="opset1">
+		<layer id="3628" name="Convert_74624" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -93704,8 +55067,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6125" name="self.model.gpt_neox.layers.25.mlp.dense_h_to_4h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 1" offset="2128922834" size="10240" />
+		<layer id="3629" name="self.model.gpt_neox.layers.22.mlp.dense_h_to_4h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 1" offset="1889139280" size="10240" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>10240</dim>
@@ -93713,7 +55076,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6126" name="Convert_22447" type="Convert" version="opset1">
+		<layer id="3630" name="Convert_74627" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -93728,7 +55091,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6127" name="self.model.gpt_neox.layers.25.mlp.dense_h_to_4h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="3631" name="self.model.gpt_neox.layers.22.mlp.dense_h_to_4h.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -93747,8 +55110,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6128" name="self.model.gpt_neox.layers.25.mlp.dense_h_to_4h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="10240, 1" offset="2128933074" size="20480" />
+		<layer id="3632" name="self.model.gpt_neox.layers.22.mlp.dense_h_to_4h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="10240, 1" offset="1889149520" size="20480" />
 			<output>
 				<port id="0" precision="FP16">
 					<dim>10240</dim>
@@ -93756,7 +55119,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6129" name="self.model.gpt_neox.layers.25.mlp.dense_h_to_4h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="3633" name="self.model.gpt_neox.layers.22.mlp.dense_h_to_4h.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -93775,7 +55138,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6130" name="self.model.gpt_neox.layers.25.mlp.dense_h_to_4h.weight" type="Convert" version="opset1">
+		<layer id="3634" name="__module.model.gpt_neox.layers.22.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
@@ -93793,7 +55156,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6131" name="__module.model.gpt_neox.layers.25.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="3635" name="__module.model.gpt_neox.layers.22.mlp.dense_h_to_4h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -93814,37 +55177,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6132" name="Constant_299321_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 10240" offset="2128953554" size="20480" />
+		<layer id="3636" name="self.model.gpt_neox.layers.22.mlp.dense_h_to_4h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="10240" offset="1889170000" size="20480" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.22.mlp.dense_h_to_4h.bias">
 					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6133" name="Constant_299321" type="Convert" version="opset1">
+		<layer id="3637" name="__module.model.gpt_neox.layers.22.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
 			</rt_info>
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16">
 					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6134" name="__module.model.gpt_neox.layers.25.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1">
+		<layer id="3638" name="__module.model.gpt_neox.layers.22.mlp.dense_h_to_4h/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -93853,20 +55210,18 @@
 					<dim>10240</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4171">
+				<port id="2" precision="FP32" names="2942">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6135" name="__module.model.gpt_neox.layers.25.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
+		<layer id="3639" name="__module.model.gpt_neox.layers.22.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
 			<data approximation_mode="ERF" />
 			<input>
 				<port id="0" precision="FP32">
@@ -93876,15 +55231,15 @@
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="4172">
+				<port id="1" precision="FP32" names="2943">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6136" name="self.model.gpt_neox.layers.25.mlp.dense_4h_to_h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 10240" offset="2128974034" size="26214400" />
+		<layer id="3640" name="self.model.gpt_neox.layers.22.mlp.dense_4h_to_h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 10240" offset="1889190480" size="26214400" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>2560</dim>
@@ -93892,7 +55247,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6137" name="Convert_22454" type="Convert" version="opset1">
+		<layer id="3641" name="Convert_74634" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -93907,8 +55262,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6138" name="self.model.gpt_neox.layers.25.mlp.dense_4h_to_h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="2155188434" size="2560" />
+		<layer id="3642" name="self.model.gpt_neox.layers.22.mlp.dense_4h_to_h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="1915404880" size="2560" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>2560</dim>
@@ -93916,7 +55271,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6139" name="Convert_22457" type="Convert" version="opset1">
+		<layer id="3643" name="Convert_74637" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -93931,7 +55286,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6140" name="self.model.gpt_neox.layers.25.mlp.dense_4h_to_h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="3644" name="self.model.gpt_neox.layers.22.mlp.dense_4h_to_h.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -93950,8 +55305,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6141" name="self.model.gpt_neox.layers.25.mlp.dense_4h_to_h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="2155190994" size="5120" />
+		<layer id="3645" name="self.model.gpt_neox.layers.22.mlp.dense_4h_to_h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="1915407440" size="5120" />
 			<output>
 				<port id="0" precision="FP16">
 					<dim>2560</dim>
@@ -93959,7 +55314,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6142" name="self.model.gpt_neox.layers.25.mlp.dense_4h_to_h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="3646" name="self.model.gpt_neox.layers.22.mlp.dense_4h_to_h.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -93978,7 +55333,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6143" name="self.model.gpt_neox.layers.25.mlp.dense_4h_to_h.weight" type="Convert" version="opset1">
+		<layer id="3647" name="__module.model.gpt_neox.layers.22.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
@@ -93996,7 +55351,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6144" name="__module.model.gpt_neox.layers.25.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="3648" name="__module.model.gpt_neox.layers.22.mlp.dense_4h_to_h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -94017,37 +55372,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6145" name="Constant_299322_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2155196114" size="5120" />
+		<layer id="3649" name="self.model.gpt_neox.layers.22.mlp.dense_4h_to_h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="2560" offset="1915412560" size="5120" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.22.mlp.dense_4h_to_h.bias">
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6146" name="Constant_299322" type="Convert" version="opset1">
+		<layer id="3650" name="__module.model.gpt_neox.layers.22.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
 			</rt_info>
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16">
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6147" name="__module.model.gpt_neox.layers.25.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1">
+		<layer id="3651" name="__module.model.gpt_neox.layers.22.mlp.dense_4h_to_h/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -94056,50 +55405,28 @@
 					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4175,input.105">
+				<port id="2" precision="FP32" names="2946,input.91">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6148" name="Constant_299323_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2155201234" size="5120" />
+		<layer id="3652" name="Constant_37466" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1915417680" size="10240" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6149" name="Constant_299323" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6150" name="__module.model.gpt_neox.layers.25.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
+		<layer id="3653" name="__module.model.gpt_neox.layers.22.input_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -94121,37 +55448,17 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6151" name="Constant_299324_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2155206354" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6152" name="Constant_299324" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="3654" name="Constant_37467" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1915427920" size="10240" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6153" name="__module.model.gpt_neox.layers.25.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+		<layer id="3655" name="__module.model.gpt_neox.layers.22.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -94166,15 +55473,15 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4041">
+				<port id="2" precision="FP32" names="2849,hidden_states.89">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6154" name="self.model.gpt_neox.layers.25.attention.query_key_value.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 2560" offset="2155211474" size="19660800" />
+		<layer id="3656" name="self.model.gpt_neox.layers.22.attention.query_key_value.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 2560" offset="1915438160" size="19660800" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>7680</dim>
@@ -94182,7 +55489,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6155" name="Convert_22424" type="Convert" version="opset1">
+		<layer id="3657" name="Convert_75066" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -94197,8 +55504,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6156" name="self.model.gpt_neox.layers.25.attention.query_key_value.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 1" offset="2174872274" size="7680" />
+		<layer id="3658" name="self.model.gpt_neox.layers.22.attention.query_key_value.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 1" offset="1935098960" size="7680" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>7680</dim>
@@ -94206,7 +55513,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6157" name="Convert_22427" type="Convert" version="opset1">
+		<layer id="3659" name="Convert_75069" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -94221,7 +55528,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6158" name="self.model.gpt_neox.layers.25.attention.query_key_value.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="3660" name="self.model.gpt_neox.layers.22.attention.query_key_value.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -94240,8 +55547,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6159" name="self.model.gpt_neox.layers.25.attention.query_key_value.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="7680, 1" offset="2174879954" size="15360" />
+		<layer id="3661" name="self.model.gpt_neox.layers.22.attention.query_key_value.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="7680, 1" offset="1935106640" size="15360" />
 			<output>
 				<port id="0" precision="FP16">
 					<dim>7680</dim>
@@ -94249,7 +55556,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6160" name="self.model.gpt_neox.layers.25.attention.query_key_value.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="3662" name="self.model.gpt_neox.layers.22.attention.query_key_value.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -94268,11 +55575,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6161" name="self.model.gpt_neox.layers.25.attention.query_key_value.weight" type="Convert" version="opset1">
+		<layer id="3663" name="self.model.gpt_neox.layers.22.attention.query_key_value.weight/fq_weights_1/convert" type="Convert" version="opset1">
 			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
 			<input>
 				<port id="0" precision="FP16">
 					<dim>7680</dim>
@@ -94286,7 +55590,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6162" name="__module.model.gpt_neox.layers.25.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="3664" name="__module.model.gpt_neox.layers.22.attention.query_key_value/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -94307,37 +55611,17 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6163" name="Constant_299325_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 7680" offset="2174895314" size="15360" />
+		<layer id="3665" name="Constant_37468" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 7680" offset="1935122000" size="30720" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6164" name="Constant_299325" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6165" name="__module.model.gpt_neox.layers.25.attention.query_key_value/aten::linear/Add" type="Add" version="opset1">
+		<layer id="3666" name="__module.model.gpt_neox.layers.22.attention.query_key_value/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -94352,25 +55636,22 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4048,qkv.101">
+				<port id="2" precision="FP32" names="2856,qkv.89">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6166" name="__module.model.gpt_neox.layers.25.attention/prim::ListConstruct/Concat" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080154" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3667" name="__module.model.gpt_neox.layers.22.attention/prim::ListConstruct" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201115992" size="32" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6167" name="__module.model.gpt_neox.layers.25.attention/aten::view/Reshape" type="Reshape" version="opset1">
+		<layer id="3668" name="__module.model.gpt_neox.layers.22.attention/aten::view/Reshape" type="Reshape" version="opset1">
 			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -94383,7 +55664,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4052,qkv.103">
+				<port id="2" precision="FP32" names="2860,qkv.91">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -94391,41 +55672,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6168" name="Constant_278404" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3669" name="Constant_14576" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6169" name="Constant_278407" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3670" name="Constant_14578" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6170" name="Constant_278410" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3671" name="Constant_14580" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6171" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3672" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Slice" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -94434,17 +55705,20 @@
 					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4053">
+				<port id="5" precision="FP32" names="2861">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -94452,15 +55726,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6172" name="Constant_14748" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="3673" name="Constant_14589" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64" names="4054">
+				<port id="0" precision="I64" names="2862">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6173" name="__module.model.gpt_neox.layers.25.attention/aten::permute/Transpose" type="Transpose" version="opset1">
+		<layer id="3674" name="__module.model.gpt_neox.layers.22.attention/aten::permute/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -94473,7 +55747,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4055,query.151">
+				<port id="2" precision="FP32" names="2863,query.45">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -94481,41 +55755,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6174" name="Constant_278416" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3675" name="Constant_14630" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6175" name="Constant_278419" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3676" name="Constant_14632" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6176" name="Constant_278422" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3677" name="Constant_14634" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6177" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3678" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Slice_1" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -94524,17 +55788,20 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4062,q.51">
+				<port id="5" precision="FP32" names="2870,q.45">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -94542,219 +55809,152 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6178" name="self.model.gpt_neox.layers.25.attention.rotary_emb.cos_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201080346" size="81920" />
-			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.25.attention.rotary_emb.cos_cached">
-					<dim>2048</dim>
+		<layer id="3679" name="__module.model.gpt_neox.layers.22.attention/aten::mul/Multiply" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="6179" name="self.model.gpt_neox.layers.25.attention.rotary_emb.cos_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
+				<port id="2" precision="FP32" names="2876">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6180" name="__module.model.gpt_neox.layers.25.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3680" name="Constant_14788" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6181" name="Constant_278428" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3681" name="Constant_14790" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6182" name="Constant_278431" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6183" name="Constant_278434" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3682" name="Constant_14792" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6184" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3683" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Slice_2" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>240</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4056">
-					<dim>-1</dim>
+				<port id="5" precision="FP32" names="2886,x2.89">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6185" name="Constant_14762" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="3684" name="Constant_37469" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="0" precision="I64" names="4057">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6186" name="__module.model.gpt_neox.layers.25.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
+		<layer id="3685" name="__module.model.gpt_neox.layers.22.attention/aten::neg/Multiply" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4058,key.151">
+				<port id="2" precision="FP32" names="2887">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6187" name="__module.model.gpt_neox.layers.25.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6188" name="__module.model.gpt_neox.layers.25.attention/aten::size/Constant_2" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="6189" name="__module.model.gpt_neox.layers.25.attention/aten::size/Gather_2" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="4066" />
-			</output>
-		</layer>
-		<layer id="6190" name="Constant_338079_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="6191" name="Constant_338079" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="6192" name="Constant_338072" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
+		<layer id="3686" name="Constant_14685" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6193" name="Constant_338074" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
+		<layer id="3687" name="Constant_14729" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6194" name="Constant_338076" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
+		<layer id="3688" name="Constant_14731" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6195" name="Concat_338077" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="3689" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Slice_3" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -94765,163 +55965,125 @@
 				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
 				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6196" name="Broadcast_338080" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="5" precision="FP32" names="2881">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6197" name="ReadValue_330497" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.25.keypresent.25.key" variable_type="f32" variable_shape="?,32,?,80" />
+		<layer id="3690" name="__module.model.gpt_neox.layers.22.attention/aten::cat/Concat" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="past_key_values.25.key">
+				<port id="2" precision="FP32" names="2889">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6198" name="Constant_323535" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64" />
-			</output>
-		</layer>
-		<layer id="6199" name="Gather_323536" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="3691" name="__module.model.gpt_neox.layers.22.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="I32">
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
-				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="3" precision="FP32">
+				<port id="2" precision="FP32" names="2890">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6200" name="__module.model.gpt_neox.layers.25.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="3692" name="__module.model.gpt_neox.layers.22.attention/aten::add/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="2891">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6201" name="__module.model.gpt_neox.layers.25.attention/aten::size/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3693" name="Constant_14640" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="6202" name="__module.model.gpt_neox.layers.25.attention/aten::size/Gather_3" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="4068" />
-			</output>
-		</layer>
-		<layer id="6203" name="__module.model.gpt_neox.layers.25.attention/aten::add_/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" names="4067,4071,4072,seq_len.101" />
-			</output>
-		</layer>
-		<layer id="6204" name="__module.model.gpt_neox.layers.25.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
 			</output>
 		</layer>
-		<layer id="6205" name="__module.model.gpt_neox.layers.25.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="3694" name="Constant_14642" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6206" name="__module.model.gpt_neox.layers.25.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3695" name="Constant_14644" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6207" name="__module.model.gpt_neox.layers.25.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3696" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Slice_4" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -94932,197 +56094,192 @@
 				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4075,4078,cos.101">
+				<port id="5" precision="FP32" names="2871">
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6208" name="Constant_38347" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="6209" name="__module.model.gpt_neox.layers.25.attention/aten::index/Gather" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="3697" name="__module.model.gpt_neox.layers.22.attention/aten::cat/Concat_1" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
-				<port id="1" precision="I32">
+				<port id="1" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="4081">
+				<port id="2" precision="FP32" names="2909">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6210" name="__module.model.gpt_neox.layers.25.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64" />
-			</input>
+		<layer id="3698" name="Constant_53079" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="2" precision="FP32" names="4082,cos.103">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
+				<port id="0" precision="FP32" />
 			</output>
 		</layer>
-		<layer id="6211" name="__module.model.gpt_neox.layers.25.attention/aten::mul/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
+		<layer id="3699" name="Constant_53072" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
-				<port id="2" precision="FP32" names="4086">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6212" name="Constant_278439" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3700" name="Constant_53074" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6213" name="Constant_278438" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3701" name="Constant_53076" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6214" name="__module.model.gpt_neox.layers.25.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="3702" name="Concat_53077" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
+				<port id="4" precision="I64">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6215" name="__module.model.gpt_neox.layers.25.attention/aten::size/Constant_5" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="6216" name="__module.model.gpt_neox.layers.25.attention/aten::size/Gather_5" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="3703" name="Broadcast_53080" type="Broadcast" version="opset3">
+			<data mode="numpy" />
 			<input>
-				<port id="0" precision="I64">
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="4087,4092" />
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6217" name="__module.model.gpt_neox.layers.25.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
+		<layer id="3704" name="ReadValue_50243" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.22.keypresent.22.key" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64" />
+				<port id="1" precision="FP32" names="past_key_values.22.key">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3705" name="Constant_47910" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="6218" name="__module.model.gpt_neox.layers.25.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1">
+		<layer id="3706" name="Gather_47911" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
-				<port id="0" precision="I64" />
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I32">
+					<dim>-1</dim>
+				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="1" precision="I64" names="4089,4090,4094,4095" />
+				<port id="3" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6219" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Constant_1" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3707" name="Constant_14594" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6220" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="3708" name="Constant_14596" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6221" name="Constant_278437" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3709" name="Constant_14598" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6222" name="ScatterUpdate_278440" type="ScatterUpdate" version="opset3">
+		<layer id="3710" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Slice_5" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -95130,148 +56287,155 @@
 				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
 				<port id="4" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="6223" name="Constant_278443" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="5" precision="FP32" names="2864">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6224" name="Constant_278446" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3711" name="Constant_14607" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="I64" names="2865">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6225" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3712" name="__module.model.gpt_neox.layers.22.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4096,x2.101">
+				<port id="2" precision="FP32" names="2866,key.89">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6226" name="Constant_299326_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
+		<layer id="3713" name="Constant_14650" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3714" name="Constant_14652" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3715" name="Constant_14654" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6227" name="Constant_299326" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="3716" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Slice_6" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="5" precision="FP32" names="2872,k.45">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6228" name="__module.model.gpt_neox.layers.25.attention/aten::neg/Multiply" type="Multiply" version="opset1">
+		<layer id="3717" name="__module.model.gpt_neox.layers.22.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
+					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4097">
+				<port id="2" precision="FP32" names="2892">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6229" name="Constant_278491" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3718" name="Constant_14931" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6230" name="Constant_278496" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3719" name="Constant_14933" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6231" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3720" name="Constant_14935" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3721" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Slice_7" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -95280,17 +56444,20 @@
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4091">
+				<port id="5" precision="FP32" names="2902,x2.91">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -95298,8 +56465,19 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6232" name="__module.model.gpt_neox.layers.25.attention/aten::cat/Concat" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="3722" name="Constant_37470" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3723" name="__module.model.gpt_neox.layers.22.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -95308,75 +56486,51 @@
 					<dim>10</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4099">
+				<port id="2" precision="FP32" names="2903">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6233" name="self.model.gpt_neox.layers.25.attention.rotary_emb.sin_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201162356" size="81920" />
-			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.25.attention.rotary_emb.sin_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6234" name="self.model.gpt_neox.layers.25.attention.rotary_emb.sin_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</input>
+		<layer id="3724" name="Constant_14828" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6235" name="__module.model.gpt_neox.layers.25.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3725" name="Constant_14872" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6236" name="__module.model.gpt_neox.layers.25.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3726" name="Constant_14874" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6237" name="__module.model.gpt_neox.layers.25.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3727" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Slice_8" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>2048</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
@@ -95388,60 +56542,45 @@
 				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4076,4079,sin.101">
+				<port id="5" precision="FP32" names="2897">
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6238" name="Constant_38350" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="6239" name="__module.model.gpt_neox.layers.25.attention/aten::index/Gather_1" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="3728" name="__module.model.gpt_neox.layers.22.attention/aten::cat/Concat_2" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="4084">
-					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>10</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="6240" name="__module.model.gpt_neox.layers.25.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
+				<port id="1" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>10</dim>
 				</port>
-				<port id="1" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4085,sin.103">
+				<port id="2" precision="FP32" names="2905">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6241" name="__module.model.gpt_neox.layers.25.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
+		<layer id="3729" name="__module.model.gpt_neox.layers.22.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -95458,7 +56597,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4100">
+				<port id="2" precision="FP32" names="2906">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -95466,7 +56605,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6242" name="__module.model.gpt_neox.layers.25.attention/aten::add/Add" type="Add" version="opset1">
+		<layer id="3730" name="__module.model.gpt_neox.layers.22.attention/aten::add/Add_1" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -95483,7 +56622,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4101">
+				<port id="2" precision="FP32" names="2907">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -95491,41 +56630,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6243" name="Constant_278542" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3731" name="Constant_14660" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6244" name="Constant_278545" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3732" name="Constant_14662" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6245" name="Constant_278548" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3733" name="Constant_14664" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6246" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3734" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Slice_9" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -95534,17 +56663,20 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4063">
+				<port id="5" precision="FP32" names="2873">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -95552,7 +56684,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6247" name="__module.model.gpt_neox.layers.25.attention/aten::cat/Concat_1" type="Concat" version="opset1">
+		<layer id="3735" name="__module.model.gpt_neox.layers.22.attention/aten::cat/Concat_3" type="Concat" version="opset1">
 			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
@@ -95569,7 +56701,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4119,4129,query.153,query.155">
+				<port id="2" precision="FP32" names="2911">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -95577,41 +56709,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6248" name="Constant_278554" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6249" name="Constant_278557" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6250" name="Constant_278560" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6251" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3736" name="__module.model.gpt_neox.layers.22.attention/aten::cat/Concat_4" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -95619,163 +56718,57 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="4064,k.51">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6252" name="__module.model.gpt_neox.layers.25.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4102">
+				<port id="2" precision="FP32" names="present.22.key">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6253" name="Constant_278564" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3737" name="Constant_53091" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
+				<port id="0" precision="FP32" />
 			</output>
 		</layer>
-		<layer id="6254" name="Constant_278563" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3738" name="Constant_53084" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6255" name="__module.model.gpt_neox.layers.25.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6256" name="__module.model.gpt_neox.layers.25.attention/aten::size/Constant_7" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3739" name="Constant_53086" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="6257" name="__module.model.gpt_neox.layers.25.attention/aten::size/Gather_7" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="4103,4108" />
-			</output>
-		</layer>
-		<layer id="6258" name="__module.model.gpt_neox.layers.25.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" />
-			</output>
-		</layer>
-		<layer id="6259" name="__module.model.gpt_neox.layers.25.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-			</input>
-			<output>
-				<port id="1" precision="I64" names="4105,4106,4110,4111" />
-			</output>
-		</layer>
-		<layer id="6260" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="6261" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6262" name="Constant_278562" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3740" name="Constant_53088" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6263" name="ScatterUpdate_278565" type="ScatterUpdate" version="opset3">
+		<layer id="3741" name="Concat_53089" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -95783,7 +56776,7 @@
 				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
@@ -95793,318 +56786,318 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6264" name="Constant_278568" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
+		<layer id="3742" name="Broadcast_53092" type="Broadcast" version="opset3">
+			<data mode="numpy" />
+			<input>
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="6265" name="Constant_278571" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6266" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3743" name="ReadValue_50245" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.22.valuepresent.22.value" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4112,x2.103">
+				<port id="1" precision="FP32" names="past_key_values.22.value">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6267" name="Constant_299327_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6268" name="Constant_299327" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
+		<layer id="3744" name="Constant_47913" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="6269" name="__module.model.gpt_neox.layers.25.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3745" name="Gather_47914" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="1" precision="I32">
+					<dim>-1</dim>
 				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4113">
+				<port id="3" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6270" name="Constant_278616" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3746" name="Constant_14612" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6271" name="Constant_278621" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3747" name="Constant_14614" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3748" name="Constant_14616" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6272" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3749" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Slice_10" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4107">
+				<port id="5" precision="FP32" names="2867">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6273" name="__module.model.gpt_neox.layers.25.attention/aten::cat/Concat_2" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="3750" name="Constant_14625" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
+			<output>
+				<port id="0" precision="I64" names="2868">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3751" name="__module.model.gpt_neox.layers.22.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4115">
+				<port id="2" precision="FP32" names="2869">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6274" name="__module.model.gpt_neox.layers.25.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3752" name="__module.model.gpt_neox.layers.22.attention/aten::cat/Concat_5" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4116">
+				<port id="2" precision="FP32" names="present.22.value">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6275" name="__module.model.gpt_neox.layers.25.attention/aten::add/Add_1" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3753" name="Constant_15007" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3754" name="ShapeOf_40274" type="ShapeOf" version="opset3">
+			<data output_type="i64" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4117">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6276" name="Constant_278666" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3755" name="Constant_40275" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116168" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6277" name="Constant_278669" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3756" name="Constant_40276" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
+				<port id="0" precision="I64" />
+			</output>
+		</layer>
+		<layer id="3757" name="Gather_40277" type="Gather" version="opset8">
+			<data batch_dims="0" />
+			<input>
 				<port id="0" precision="I64">
 					<dim>4</dim>
 				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64" />
+			</input>
+			<output>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6278" name="Constant_278672" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3758" name="Add_40280" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="I64" names="2916">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3759" name="Constant_15010" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6279" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3760" name="__module.model.gpt_neox.layers.22.attention/aten::slice/Slice_14" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4065">
+				<port id="5" precision="FP32" names="2920">
+					<dim>-1</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>60</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6280" name="__module.model.gpt_neox.layers.25.attention/aten::cat/Concat_3" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="3761" name="__module.model.gpt_neox.layers.22.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+			<data causal="false" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>60</dim>
+					<dim>80</dim>
+				</port>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="3" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4121">
+				<port id="4" precision="FP32" names="2921,attn_output.133">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -96112,8 +57105,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6281" name="__module.model.gpt_neox.layers.25.attention/aten::to/Convert_1" type="Concat" version="opset1">
-			<data axis="-2" />
+		<layer id="3762" name="__module.model.gpt_neox.layers.22.attention/aten::transpose/Constant" type="Const" version="opset1">
+			<data element_type="i32" shape="4" offset="201116200" size="16" />
+			<output>
+				<port id="0" precision="I32">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3763" name="__module.model.gpt_neox.layers.22.attention/aten::transpose/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -96121,869 +57121,1044 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
+				<port id="1" precision="I32">
+					<dim>4</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="2922">
+					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
+					<dim>80</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3764" name="Constant_41006" type="Const" version="opset1">
+			<data element_type="i64" shape="3" offset="201116216" size="24" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>3</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3765" name="__module.model.gpt_neox.layers.22.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
+			<data special_zero="true" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>80</dim>
 				</port>
+				<port id="1" precision="I64">
+					<dim>3</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="present.25.key">
+				<port id="2" precision="FP32" names="2925">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6282" name="Constant_338092_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
+		<layer id="3766" name="self.model.gpt_neox.layers.22.attention.dense.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 2560" offset="1935152720" size="6553600" />
 			<output>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6283" name="Constant_338092" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="3767" name="Convert_75418" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" />
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6284" name="Constant_338085" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
+		<layer id="3768" name="self.model.gpt_neox.layers.22.attention.dense.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="1941706320" size="2560" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="U8">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6285" name="Constant_338087" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64">
+		<layer id="3769" name="Convert_75421" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="6286" name="Constant_338089" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
+			</input>
 			<output>
-				<port id="0" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6287" name="Concat_338090" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="3770" name="self.model.gpt_neox.layers.22.attention.dense.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3771" name="self.model.gpt_neox.layers.22.attention.dense.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="1941708880" size="5120" />
+			<output>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I64">
+			</output>
+		</layer>
+		<layer id="3772" name="self.model.gpt_neox.layers.22.attention.dense.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6288" name="Broadcast_338093" type="Broadcast" version="opset3">
-			<data mode="numpy" />
+		<layer id="3773" name="self.model.gpt_neox.layers.22.attention.dense.weight/fq_weights_1/convert" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6289" name="ReadValue_330499" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.25.valuepresent.25.value" variable_type="f32" variable_shape="?,32,?,80" />
+		<layer id="3774" name="__module.model.gpt_neox.layers.22.attention.dense/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="past_key_values.25.value">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6290" name="Constant_323538" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
+		<layer id="3775" name="Constant_37471" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1941714000" size="10240" />
 			<output>
-				<port id="0" precision="I64" />
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6291" name="Gather_323539" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="3776" name="__module.model.gpt_neox.layers.22.attention.dense/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="3" precision="FP32">
+				<port id="2" precision="FP32" names="2928,2930,input.89">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6292" name="Constant_278678" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+		<layer id="3777" name="__module.model.gpt_neox.layers.22/aten::add/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="6293" name="Constant_278681" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="6294" name="Constant_278684" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="2948">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6295" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3778" name="__module.model.gpt_neox.layers.22/aten::add/Add_1" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4059">
+				<port id="2" precision="FP32" names="2949,2951,hidden_states.95">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6296" name="Constant_14776" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="3779" name="__module.model.gpt_neox.layers.23.input_layernorm/aten::layer_norm/Multiply" type="Const" version="opset1">
+			<data element_type="i32" shape="1" offset="128867668" size="4" />
 			<output>
-				<port id="0" precision="I64" names="4060">
-					<dim>4</dim>
+				<port id="0" precision="I32">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6297" name="__module.model.gpt_neox.layers.25.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
+		<layer id="3780" name="__module.model.gpt_neox.layers.23.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
+			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="I32">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4061">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6298" name="__module.model.gpt_neox.layers.25.attention/aten::cat/Concat_5" type="Concat" version="opset1">
-			<data axis="-2" />
+		<layer id="3781" name="Constant_37472" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1941724240" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3782" name="__module.model.gpt_neox.layers.23.post_attention_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="present.25.value">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6299" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Slice_11" type="Const" version="opset1">
-			<data element_type="boolean" shape="1, 1, 2048, 2048" offset="201244276" size="4194304" />
+		<layer id="3783" name="Constant_37473" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1941734480" size="10240" />
 			<output>
-				<port id="0" precision="BOOL" names="4138">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6300" name="Constant_278688" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3784" name="__module.model.gpt_neox.layers.23.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="2" precision="FP32" names="3049">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6301" name="Constant_278687" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162298" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3785" name="self.model.gpt_neox.layers.23.mlp.dense_h_to_4h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 2560" offset="1941744720" size="26214400" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6302" name="__module.model.gpt_neox.layers.25.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="3786" name="Convert_74644" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6303" name="__module.model.gpt_neox.layers.25.attention/aten::size/Constant_8" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3787" name="self.model.gpt_neox.layers.23.mlp.dense_h_to_4h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 1" offset="1967959120" size="10240" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6304" name="__module.model.gpt_neox.layers.25.attention/aten::size/Gather_8" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="3788" name="Convert_74647" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="4133" />
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6305" name="__module.model.gpt_neox.layers.25.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="3789" name="self.model.gpt_neox.layers.23.mlp.dense_h_to_4h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6306" name="__module.model.gpt_neox.layers.25.attention/aten::size/Constant_9" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3790" name="self.model.gpt_neox.layers.23.mlp.dense_h_to_4h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="10240, 1" offset="1967969360" size="20480" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="6307" name="__module.model.gpt_neox.layers.25.attention/aten::size/Gather_9" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="4131" />
 			</output>
 		</layer>
-		<layer id="6308" name="__module.model.gpt_neox.layers.25.attention/aten::sub/Subtract" type="Subtract" version="opset1">
+		<layer id="3791" name="self.model.gpt_neox.layers.23.mlp.dense_h_to_4h.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64" names="4135,4136" />
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6309" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Constant_4" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
+		<layer id="3792" name="__module.model.gpt_neox.layers.23.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<rt_info>
-				<attribute name="precise" version="0" />
+				<attribute name="decompression" version="0" />
 			</rt_info>
+			<input>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I32" />
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6310" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1">
+		<layer id="3793" name="__module.model.gpt_neox.layers.23.mlp.dense_h_to_4h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3794" name="self.model.gpt_neox.layers.23.mlp.dense_h_to_4h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="10240" offset="1967989840" size="20480" />
+			<output>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.23.mlp.dense_h_to_4h.bias">
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6311" name="Constant_278686" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
+		<layer id="3795" name="__module.model.gpt_neox.layers.23.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<rt_info>
-				<attribute name="precise" version="0" />
+				<attribute name="decompression" version="0" />
 			</rt_info>
+			<input>
+				<port id="0" precision="BF16">
+					<dim>10240</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6312" name="ScatterUpdate_278689" type="ScatterUpdate" version="opset3">
+		<layer id="3796" name="__module.model.gpt_neox.layers.23.mlp.dense_h_to_4h/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
+				<port id="2" precision="FP32" names="3054">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6313" name="Constant_278690" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3797" name="__module.model.gpt_neox.layers.23.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
+			<data approximation_mode="ERF" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="1" precision="FP32" names="3055">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6314" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="3798" name="self.model.gpt_neox.layers.23.mlp.dense_4h_to_h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 10240" offset="1968010320" size="26214400" />
 			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6315" name="ScatterUpdate_278691" type="ScatterUpdate" version="opset3">
+		<layer id="3799" name="Convert_74654" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6316" name="Constant_278694" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438612" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3800" name="self.model.gpt_neox.layers.23.mlp.dense_4h_to_h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="1994224720" size="2560" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6317" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3801" name="Convert_74657" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
 					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>3</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="BOOL" names="4139">
-					<dim>1</dim>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6318" name="Constant_278767" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+		<layer id="3802" name="self.model.gpt_neox.layers.23.mlp.dense_4h_to_h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="6319" name="Constant_278768" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="6320" name="Constant_278764" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6321" name="Constant_278763" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3803" name="self.model.gpt_neox.layers.23.mlp.dense_4h_to_h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="1994227280" size="5120" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6322" name="ScatterUpdate_278769" type="ScatterUpdate" version="opset3">
+		<layer id="3804" name="self.model.gpt_neox.layers.23.mlp.dense_4h_to_h.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6323" name="Constant_278772" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
+		<layer id="3805" name="__module.model.gpt_neox.layers.23.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<rt_info>
-				<attribute name="precise" version="0" />
+				<attribute name="decompression" version="0" />
 			</rt_info>
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6324" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3806" name="__module.model.gpt_neox.layers.23.mlp.dense_4h_to_h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="BOOL" names="4140,4141">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6325" name="__module.model.gpt_neox.layers.25.attention/aten::where/Convert_1_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
+		<layer id="3807" name="self.model.gpt_neox.layers.23.mlp.dense_4h_to_h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="2560" offset="1994232400" size="5120" />
 			<output>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.23.mlp.dense_4h_to_h.bias">
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6326" name="__module.model.gpt_neox.layers.25.attention/aten::where/Convert_1" type="Convert" version="opset1">
+		<layer id="3808" name="__module.model.gpt_neox.layers.23.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
 			</rt_info>
 			<input>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="BF16">
+					<dim>2560</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="6327" name="__module.model.gpt_neox.layers.25.attention/aten::full/Convert" type="Const" version="opset1">
-			<data element_type="f32" shape="" offset="205438636" size="4" />
-			<output>
-				<port id="0" precision="FP32" names="4128" />
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6328" name="__module.model.gpt_neox.layers.25.attention/aten::where/Select" type="Select" version="opset1">
+		<layer id="3809" name="__module.model.gpt_neox.layers.23.mlp.dense_4h_to_h/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="FP32" />
-				<port id="2" precision="FP32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="4142,causal_mask.101">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="3058,input.95">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6329" name="Add_290488" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
+		<layer id="3810" name="Constant_37474" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1994237520" size="10240" />
+			<output>
 				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3811" name="__module.model.gpt_neox.layers.23.input_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6330" name="ShapeOf_290489" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="3812" name="Constant_37475" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="1994247760" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3813" name="__module.model.gpt_neox.layers.23.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
 					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="2961,hidden_states.93">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3814" name="self.model.gpt_neox.layers.23.attention.query_key_value.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 2560" offset="1994258000" size="19660800" />
+			<output>
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3815" name="Convert_75077" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6331" name="Constant_254704" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3816" name="self.model.gpt_neox.layers.23.attention.query_key_value.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 1" offset="2013918800" size="7680" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="U8">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6332" name="__module.model.gpt_neox.layers.25.attention/aten::size/Constant_10" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3817" name="Convert_75080" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>1</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I32" />
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6333" name="__module.model.gpt_neox.layers.25.attention/aten::size/Gather_10" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="3818" name="self.model.gpt_neox.layers.23.attention.query_key_value.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="4126">
+				<port id="2" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3819" name="self.model.gpt_neox.layers.23.attention.query_key_value.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="7680, 1" offset="2013926480" size="15360" />
+			<output>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6334" name="__module.model.gpt_neox.layers.25.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="3820" name="self.model.gpt_neox.layers.23.attention.query_key_value.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
+			</output>
+		</layer>
+		<layer id="3821" name="self.model.gpt_neox.layers.23.attention.query_key_value.weight/fq_weights_1/convert" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64" names="4143">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6335" name="__module.model.gpt_neox.layers.25.attention/aten::expand/Abs" type="Abs" version="opset1">
+		<layer id="3822" name="__module.model.gpt_neox.layers.23.attention.query_key_value/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>7680</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3823" name="Constant_37476" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 7680" offset="2013941840" size="30720" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6336" name="Maximum_290490" type="Maximum" version="opset1">
+		<layer id="3824" name="__module.model.gpt_neox.layers.23.attention.query_key_value/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>7680</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>7680</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
+				<port id="2" precision="FP32" names="2968,qkv.93">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>7680</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3825" name="__module.model.gpt_neox.layers.23.attention/prim::ListConstruct" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201115992" size="32" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6337" name="__module.model.gpt_neox.layers.25.attention/aten::add/Add_2" type="Broadcast" version="opset3">
-			<data mode="numpy" />
+		<layer id="3826" name="__module.model.gpt_neox.layers.23.attention/aten::view/Reshape" type="Reshape" version="opset1">
+			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>7680</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4145">
-					<dim>-1</dim>
-					<dim>-1</dim>
+				<port id="2" precision="FP32" names="2972,qkv.95">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6338" name="__module.model.gpt_neox.layers.25.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
-			<data causal="false" />
+		<layer id="3827" name="Constant_15206" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3828" name="Constant_15208" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3829" name="Constant_15210" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3830" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Slice" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>240</dim>
 				</port>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+				<port id="1" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4146,4147,sdpa_result.51,tensor.101">
+				<port id="5" precision="FP32" names="2973">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6339" name="Constant_15217" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="3831" name="Constant_15219" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64" names="4148">
+				<port id="0" precision="I64" names="2974">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6340" name="__module.model.gpt_neox.layers.25.attention/aten::permute/Transpose_3" type="Transpose" version="opset1">
+		<layer id="3832" name="__module.model.gpt_neox.layers.23.attention/aten::permute/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
@@ -96991,1154 +58166,1135 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4149">
-					<dim>-1</dim>
+				<port id="2" precision="FP32" names="2975,query.47">
 					<dim>-1</dim>
 					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6341" name="__module.model.gpt_neox.layers.25.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438650" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3833" name="Constant_15260" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>3</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6342" name="__module.model.gpt_neox.layers.25.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
-			<data special_zero="true" />
+		<layer id="3834" name="Constant_15262" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3835" name="Constant_15264" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3836" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Slice_1" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>3</dim>
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4154">
+				<port id="5" precision="FP32" names="2982,q.47">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6343" name="self.model.gpt_neox.layers.25.attention.dense.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 2560" offset="2174910674" size="6553600" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6344" name="Convert_22434" type="Convert" version="opset1">
-			<data destination_type="f16" />
+		<layer id="3837" name="__module.model.gpt_neox.layers.23.attention/aten::mul/Multiply" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32" names="2988">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6345" name="self.model.gpt_neox.layers.25.attention.dense.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="2181464274" size="2560" />
+		<layer id="3838" name="Constant_15418" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6346" name="Convert_22437" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
+		<layer id="3839" name="Constant_15420" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="3840" name="Constant_15422" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6347" name="self.model.gpt_neox.layers.25.attention.dense.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3841" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Slice_2" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6348" name="self.model.gpt_neox.layers.25.attention.dense.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="2181466834" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
+				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="6349" name="self.model.gpt_neox.layers.25.attention.dense.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="3" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="5" precision="FP32" names="2998,x2.93">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6350" name="self.model.gpt_neox.layers.25.attention.dense.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="3842" name="Constant_37477" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6351" name="__module.model.gpt_neox.layers.25.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
+		<layer id="3843" name="__module.model.gpt_neox.layers.23.attention/aten::neg/Multiply" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>2560</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="2999">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6352" name="Constant_299328_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2181471954" size="5120" />
+		<layer id="3844" name="Constant_15315" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6353" name="Constant_299328" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+		<layer id="3845" name="Constant_15359" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="3846" name="Constant_15361" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6354" name="__module.model.gpt_neox.layers.25.attention.dense/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3847" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Slice_3" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="FP32">
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4157,4159,input.103">
+				<port id="5" precision="FP32" names="2993">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6355" name="__module.model.gpt_neox.layers.25/aten::add/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3848" name="__module.model.gpt_neox.layers.23.attention/aten::cat/Concat" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4177">
+				<port id="2" precision="FP32" names="3001">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6356" name="__module.model.gpt_neox.layers.25/aten::add/Add_1" type="Add" version="opset1">
+		<layer id="3849" name="__module.model.gpt_neox.layers.23.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4178,4180,hidden_states.53">
+				<port id="2" precision="FP32" names="3002">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6357" name="__module.model.gpt_neox.layers.26.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867670" size="4" />
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6358" name="__module.model.gpt_neox.layers.26.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
-			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
+		<layer id="3850" name="__module.model.gpt_neox.layers.23.attention/aten::add/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="I32">
-					<dim>1</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="3003">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6359" name="Constant_299329_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2181477074" size="5120" />
+		<layer id="3851" name="Constant_15270" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6360" name="Constant_299329" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+		<layer id="3852" name="Constant_15272" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="3853" name="Constant_15274" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6361" name="__module.model.gpt_neox.layers.26.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3854" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Slice_4" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="1" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6362" name="Constant_299330_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2181482194" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="2" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="6363" name="Constant_299330" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
+				<port id="3" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+				<port id="5" precision="FP32" names="2983">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6364" name="__module.model.gpt_neox.layers.26.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3855" name="__module.model.gpt_neox.layers.23.attention/aten::cat/Concat_1" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4315">
+				<port id="2" precision="FP32" names="3021">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6365" name="self.model.gpt_neox.layers.26.mlp.dense_h_to_4h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 2560" offset="2181487314" size="26214400" />
+		<layer id="3856" name="Constant_53103" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
+				<port id="0" precision="FP32" />
 			</output>
 		</layer>
-		<layer id="6366" name="Convert_22484" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="3857" name="Constant_53096" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6367" name="self.model.gpt_neox.layers.26.mlp.dense_h_to_4h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 1" offset="2207701714" size="10240" />
+		<layer id="3858" name="Constant_53098" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6368" name="Convert_22487" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</input>
+		<layer id="3859" name="Constant_53100" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6369" name="self.model.gpt_neox.layers.26.mlp.dense_h_to_4h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3860" name="Concat_53101" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
+				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="2" precision="I64">
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="6370" name="self.model.gpt_neox.layers.26.mlp.dense_h_to_4h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="10240, 1" offset="2207711954" size="20480" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
+			</input>
+			<output>
+				<port id="4" precision="I64">
+					<dim>4</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6371" name="self.model.gpt_neox.layers.26.mlp.dense_h_to_4h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3861" name="Broadcast_53104" type="Broadcast" version="opset3">
+			<data mode="numpy" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6372" name="self.model.gpt_neox.layers.26.mlp.dense_h_to_4h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="3862" name="ReadValue_50247" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.23.keypresent.23.key" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="1" precision="FP32" names="past_key_values.23.key">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6373" name="__module.model.gpt_neox.layers.26.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
+		<layer id="3863" name="Constant_47916" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64" />
+			</output>
+		</layer>
+		<layer id="3864" name="Gather_47917" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="1" precision="I32">
+					<dim>-1</dim>
 				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="3" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6374" name="Constant_299331_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 10240" offset="2207732434" size="20480" />
+		<layer id="3865" name="Constant_15224" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6375" name="Constant_299331" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+		<layer id="3866" name="Constant_15226" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>10240</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="3867" name="Constant_15228" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6376" name="__module.model.gpt_neox.layers.26.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3868" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Slice_5" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
-				<port id="1" precision="FP32">
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
 					<dim>1</dim>
-					<dim>10240</dim>
 				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="4320">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
+				<port id="3" precision="I64">
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="6377" name="__module.model.gpt_neox.layers.26.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
-			<data approximation_mode="ERF" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="4321">
+				<port id="5" precision="FP32" names="2976">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6378" name="self.model.gpt_neox.layers.26.mlp.dense_4h_to_h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 10240" offset="2207752914" size="26214400" />
+		<layer id="3869" name="Constant_15237" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="I64" names="2977">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6379" name="Convert_22494" type="Convert" version="opset1">
-			<data destination_type="f16" />
+		<layer id="3870" name="__module.model.gpt_neox.layers.23.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
 			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="2" precision="FP32" names="2978,key.93">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6380" name="self.model.gpt_neox.layers.26.mlp.dense_4h_to_h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="2233967314" size="2560" />
+		<layer id="3871" name="Constant_15280" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6381" name="Convert_22497" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
+		<layer id="3872" name="Constant_15282" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6382" name="self.model.gpt_neox.layers.26.mlp.dense_4h_to_h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
 			</output>
 		</layer>
-		<layer id="6383" name="self.model.gpt_neox.layers.26.mlp.dense_4h_to_h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="2233969874" size="5120" />
+		<layer id="3873" name="Constant_15284" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6384" name="self.model.gpt_neox.layers.26.mlp.dense_4h_to_h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3874" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Slice_6" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6385" name="self.model.gpt_neox.layers.26.mlp.dense_4h_to_h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="5" precision="FP32" names="2984,k.47">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6386" name="__module.model.gpt_neox.layers.26.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
+		<layer id="3875" name="__module.model.gpt_neox.layers.23.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="3004">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6387" name="Constant_299332_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2233974994" size="5120" />
+		<layer id="3876" name="Constant_15561" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6388" name="Constant_299332" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+		<layer id="3877" name="Constant_15563" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="3878" name="Constant_15565" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6389" name="__module.model.gpt_neox.layers.26.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3879" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Slice_7" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="FP32">
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4324,input.109">
+				<port id="5" precision="FP32" names="3014,x2.95">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6390" name="Constant_299333_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2233980114" size="5120" />
+		<layer id="3880" name="Constant_37478" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6391" name="Constant_299333" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6392" name="__module.model.gpt_neox.layers.26.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
+		<layer id="3881" name="__module.model.gpt_neox.layers.23.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="3015">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6393" name="Constant_299334_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2233985234" size="5120" />
+		<layer id="3882" name="Constant_15458" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="FP16">
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3883" name="Constant_15502" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6394" name="Constant_299334" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="3884" name="Constant_15504" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3885" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Slice_8" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
+				<port id="3" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="5" precision="FP32" names="3009">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6395" name="__module.model.gpt_neox.layers.26.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3886" name="__module.model.gpt_neox.layers.23.attention/aten::cat/Concat_2" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4190">
+				<port id="2" precision="FP32" names="3017">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6396" name="self.model.gpt_neox.layers.26.attention.query_key_value.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 2560" offset="2233990354" size="19660800" />
+		<layer id="3887" name="__module.model.gpt_neox.layers.23.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32" names="3018">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6397" name="Convert_22464" type="Convert" version="opset1">
-			<data destination_type="f16" />
+		<layer id="3888" name="__module.model.gpt_neox.layers.23.attention/aten::add/Add_1" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32" names="3019">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6398" name="self.model.gpt_neox.layers.26.attention.query_key_value.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 1" offset="2253651154" size="7680" />
+		<layer id="3889" name="Constant_15290" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6399" name="Convert_22467" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
+		<layer id="3890" name="Constant_15292" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="3891" name="Constant_15294" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6400" name="self.model.gpt_neox.layers.26.attention.query_key_value.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3892" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Slice_9" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
+				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6401" name="self.model.gpt_neox.layers.26.attention.query_key_value.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="7680, 1" offset="2253658834" size="15360" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
+				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="6402" name="self.model.gpt_neox.layers.26.attention.query_key_value.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="3" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="5" precision="FP32" names="2985">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6403" name="self.model.gpt_neox.layers.26.attention.query_key_value.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="3893" name="__module.model.gpt_neox.layers.23.attention/aten::cat/Concat_3" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32" names="3023">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6404" name="__module.model.gpt_neox.layers.26.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
+		<layer id="3894" name="__module.model.gpt_neox.layers.23.attention/aten::cat/Concat_4" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>7680</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="present.23.key">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6405" name="Constant_299335_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 7680" offset="2253674194" size="15360" />
+		<layer id="3895" name="Constant_53115" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32" />
+			</output>
+		</layer>
+		<layer id="3896" name="Constant_53108" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3897" name="Constant_53110" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6406" name="Constant_299335" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="3898" name="Constant_53112" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3899" name="Concat_53113" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>1</dim>
-					<dim>7680</dim>
 				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
+				<port id="2" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
 					<dim>1</dim>
-					<dim>7680</dim>
+				</port>
+			</input>
+			<output>
+				<port id="4" precision="I64">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6407" name="__module.model.gpt_neox.layers.26.attention.query_key_value/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3900" name="Broadcast_53116" type="Broadcast" version="opset3">
+			<data mode="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
+					<dim>4</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
+			</output>
+		</layer>
+		<layer id="3901" name="ReadValue_50249" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.23.valuepresent.23.value" variable_type="f32" variable_shape="?,32,?,80" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4197,qkv.105">
+				<port id="1" precision="FP32" names="past_key_values.23.value">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6408" name="__module.model.gpt_neox.layers.26.attention/prim::ListConstruct/Concat" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080154" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3902" name="Constant_47919" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="6409" name="__module.model.gpt_neox.layers.26.attention/aten::view/Reshape" type="Reshape" version="opset1">
-			<data special_zero="true" />
+		<layer id="3903" name="Gather_47920" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="I32">
+					<dim>-1</dim>
 				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4201,qkv.107">
-					<dim>-1</dim>
+				<port id="3" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>240</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6410" name="Constant_278779" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3904" name="Constant_15242" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6411" name="Constant_278782" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3905" name="Constant_15244" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6412" name="Constant_278785" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3906" name="Constant_15246" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6413" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3907" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Slice_10" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -98147,17 +59303,20 @@
 					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4202">
+				<port id="5" precision="FP32" names="2979">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -98165,15 +59324,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6414" name="Constant_15329" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="3908" name="Constant_15255" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64" names="4203">
+				<port id="0" precision="I64" names="2980">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6415" name="__module.model.gpt_neox.layers.26.attention/aten::permute/Transpose" type="Transpose" version="opset1">
+		<layer id="3909" name="__module.model.gpt_neox.layers.23.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -98186,7 +59345,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4204,query.157">
+				<port id="2" precision="FP32" names="2981">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -98194,41 +59353,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6416" name="Constant_278791" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6417" name="Constant_278794" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6418" name="Constant_278797" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6419" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3910" name="__module.model.gpt_neox.layers.23.attention/aten::cat/Concat_5" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -98236,1053 +59362,1230 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4211,q.53">
+				<port id="2" precision="FP32" names="present.23.value">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6420" name="self.model.gpt_neox.layers.26.attention.rotary_emb.cos_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201080346" size="81920" />
+		<layer id="3911" name="Constant_15637" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.26.attention.rotary_emb.cos_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6421" name="self.model.gpt_neox.layers.26.attention.rotary_emb.cos_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="3912" name="ShapeOf_40345" type="ShapeOf" version="opset3">
+			<data output_type="i64" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6422" name="__module.model.gpt_neox.layers.26.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3913" name="Constant_40346" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116168" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6423" name="Constant_278803" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3914" name="Constant_40347" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
+				<port id="0" precision="I64" />
+			</output>
+		</layer>
+		<layer id="3915" name="Gather_40348" type="Gather" version="opset8">
+			<data batch_dims="0" />
+			<input>
 				<port id="0" precision="I64">
 					<dim>4</dim>
 				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64" />
+			</input>
+			<output>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6424" name="Constant_278806" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
+		<layer id="3916" name="Add_40351" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="I64" names="3028">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6425" name="Constant_278809" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3917" name="Constant_15640" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6426" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3918" name="__module.model.gpt_neox.layers.23.attention/aten::slice/Slice_14" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4205">
+				<port id="5" precision="FP32" names="3032">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3919" name="__module.model.gpt_neox.layers.23.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+			<data causal="false" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="3" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
 					<dim>-1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="4" precision="FP32" names="3033,attn_output.139">
 					<dim>-1</dim>
 					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6427" name="Constant_15343" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="3920" name="__module.model.gpt_neox.layers.23.attention/aten::transpose/Constant" type="Const" version="opset1">
+			<data element_type="i32" shape="4" offset="201116200" size="16" />
 			<output>
-				<port id="0" precision="I64" names="4206">
+				<port id="0" precision="I32">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6428" name="__module.model.gpt_neox.layers.26.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
+		<layer id="3921" name="__module.model.gpt_neox.layers.23.attention/aten::transpose/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="I32">
 					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4207,key.157">
+				<port id="2" precision="FP32" names="3034">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6429" name="__module.model.gpt_neox.layers.26.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="3922" name="Constant_41007" type="Const" version="opset1">
+			<data element_type="i64" shape="3" offset="201116216" size="24" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>3</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3923" name="__module.model.gpt_neox.layers.23.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
+			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>80</dim>
 				</port>
+				<port id="1" precision="I64">
+					<dim>3</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="3037">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6430" name="__module.model.gpt_neox.layers.26.attention/aten::size/Constant_2" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3924" name="self.model.gpt_neox.layers.23.attention.dense.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 2560" offset="2013972560" size="6553600" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6431" name="__module.model.gpt_neox.layers.26.attention/aten::size/Gather_2" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="3925" name="Convert_75429" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="4215" />
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6432" name="Constant_338105_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
+		<layer id="3926" name="self.model.gpt_neox.layers.23.attention.dense.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="2020526160" size="2560" />
 			<output>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6433" name="Constant_338105" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="3927" name="Convert_75432" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="6434" name="Constant_338098" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
-			<output>
-				<port id="0" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6435" name="Constant_338100" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64">
+		<layer id="3928" name="self.model.gpt_neox.layers.23.attention.dense.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6436" name="Constant_338102" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
+		<layer id="3929" name="self.model.gpt_neox.layers.23.attention.dense.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="2020528720" size="5120" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6437" name="Concat_338103" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="3930" name="self.model.gpt_neox.layers.23.attention.dense.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
+			</output>
+		</layer>
+		<layer id="3931" name="self.model.gpt_neox.layers.23.attention.dense.weight/fq_weights_1/convert" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6438" name="Broadcast_338106" type="Broadcast" version="opset3">
-			<data mode="numpy" />
+		<layer id="3932" name="__module.model.gpt_neox.layers.23.attention.dense/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6439" name="ReadValue_330501" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.26.keypresent.26.key" variable_type="f32" variable_shape="?,32,?,80" />
+		<layer id="3933" name="Constant_37479" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2020533840" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3934" name="__module.model.gpt_neox.layers.23.attention.dense/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="past_key_values.26.key">
+				<port id="2" precision="FP32" names="3040,3042,input.93">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6440" name="Constant_323541" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
+		<layer id="3935" name="__module.model.gpt_neox.layers.23/aten::add/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64" />
+				<port id="2" precision="FP32" names="3060">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6441" name="Gather_323542" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="3936" name="__module.model.gpt_neox.layers.23/aten::add/Add_1" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I32">
+				<port id="1" precision="FP32">
 					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="3" precision="FP32">
+				<port id="2" precision="FP32" names="3061,3063,hidden_states.99">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6442" name="__module.model.gpt_neox.layers.26.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="3937" name="__module.model.gpt_neox.layers.24.input_layernorm/aten::layer_norm/Multiply" type="Const" version="opset1">
+			<data element_type="i32" shape="1" offset="128867668" size="4" />
+			<output>
+				<port id="0" precision="I32">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3938" name="__module.model.gpt_neox.layers.24.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
+			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="I32">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6443" name="__module.model.gpt_neox.layers.26.attention/aten::size/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3939" name="Constant_37480" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2020544080" size="10240" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6444" name="__module.model.gpt_neox.layers.26.attention/aten::size/Gather_3" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="3940" name="__module.model.gpt_neox.layers.24.post_attention_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="4217" />
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3941" name="Constant_37481" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2020554320" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6445" name="__module.model.gpt_neox.layers.26.attention/aten::add_/Add" type="Add" version="opset1">
+		<layer id="3942" name="__module.model.gpt_neox.layers.24.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64" names="4216,4220,4221,seq_len.105" />
+				<port id="2" precision="FP32" names="3161">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6446" name="__module.model.gpt_neox.layers.26.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3943" name="self.model.gpt_neox.layers.24.mlp.dense_h_to_4h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 2560" offset="2020564560" size="26214400" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6447" name="__module.model.gpt_neox.layers.26.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1">
+		<layer id="3944" name="Convert_74664" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6448" name="__module.model.gpt_neox.layers.26.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3945" name="self.model.gpt_neox.layers.24.mlp.dense_h_to_4h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 1" offset="2046778960" size="10240" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="U8">
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6449" name="__module.model.gpt_neox.layers.26.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3946" name="Convert_74667" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
+				<port id="0" precision="U8">
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
+			</input>
+			<output>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I64">
+			</output>
+		</layer>
+		<layer id="3947" name="self.model.gpt_neox.layers.24.mlp.dense_h_to_4h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4224,4227,cos.105">
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6450" name="Constant_38353" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
+		<layer id="3948" name="self.model.gpt_neox.layers.24.mlp.dense_h_to_4h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="10240, 1" offset="2046789200" size="20480" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6451" name="__module.model.gpt_neox.layers.26.attention/aten::index/Gather" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="3949" name="self.model.gpt_neox.layers.24.mlp.dense_h_to_4h.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
 				</port>
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="4230">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6452" name="__module.model.gpt_neox.layers.26.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1">
+		<layer id="3950" name="__module.model.gpt_neox.layers.24.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<rt_info>
+				<attribute name="decompression" version="0" />
+			</rt_info>
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4231,cos.107">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6453" name="__module.model.gpt_neox.layers.26.attention/aten::mul/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3951" name="__module.model.gpt_neox.layers.24.mlp.dense_h_to_4h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4235">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6454" name="Constant_278814" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3952" name="self.model.gpt_neox.layers.24.mlp.dense_h_to_4h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="10240" offset="2046809680" size="20480" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.24.mlp.dense_h_to_4h.bias">
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6455" name="Constant_278813" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
+		<layer id="3953" name="__module.model.gpt_neox.layers.24.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<rt_info>
-				<attribute name="precise" version="0" />
+				<attribute name="decompression" version="0" />
 			</rt_info>
+			<input>
+				<port id="0" precision="BF16">
+					<dim>10240</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6456" name="__module.model.gpt_neox.layers.26.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="3954" name="__module.model.gpt_neox.layers.24.mlp.dense_h_to_4h/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>10240</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="3166">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6457" name="__module.model.gpt_neox.layers.26.attention/aten::size/Constant_5" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="6458" name="__module.model.gpt_neox.layers.26.attention/aten::size/Gather_5" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="3955" name="__module.model.gpt_neox.layers.24.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
+			<data approximation_mode="ERF" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="4236,4241" />
+				<port id="1" precision="FP32" names="3167">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6459" name="__module.model.gpt_neox.layers.26.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
+		<layer id="3956" name="self.model.gpt_neox.layers.24.mlp.dense_4h_to_h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 10240" offset="2046830160" size="26214400" />
 			<output>
-				<port id="2" precision="I64" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6460" name="__module.model.gpt_neox.layers.26.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1">
+		<layer id="3957" name="Convert_74674" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64" names="4238,4239,4243,4244" />
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6461" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Constant_1" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3958" name="self.model.gpt_neox.layers.24.mlp.dense_4h_to_h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="2073044560" size="2560" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3959" name="Convert_74677" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6462" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1">
+		<layer id="3960" name="self.model.gpt_neox.layers.24.mlp.dense_4h_to_h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6463" name="Constant_278812" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3961" name="self.model.gpt_neox.layers.24.mlp.dense_4h_to_h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="2073047120" size="5120" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6464" name="ScatterUpdate_278815" type="ScatterUpdate" version="opset3">
+		<layer id="3962" name="self.model.gpt_neox.layers.24.mlp.dense_4h_to_h.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6465" name="Constant_278818" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
+		<layer id="3963" name="__module.model.gpt_neox.layers.24.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<rt_info>
-				<attribute name="precise" version="0" />
+				<attribute name="decompression" version="0" />
 			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="6466" name="Constant_278821" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6467" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3964" name="__module.model.gpt_neox.layers.24.mlp.dense_4h_to_h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4245,x2.105">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6468" name="Constant_299336_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
+		<layer id="3965" name="self.model.gpt_neox.layers.24.mlp.dense_4h_to_h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="2560" offset="2073052240" size="5120" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.24.mlp.dense_4h_to_h.bias">
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6469" name="Constant_299336" type="Convert" version="opset1">
+		<layer id="3966" name="__module.model.gpt_neox.layers.24.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
 			</rt_info>
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16">
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6470" name="__module.model.gpt_neox.layers.26.attention/aten::neg/Multiply" type="Multiply" version="opset1">
+		<layer id="3967" name="__module.model.gpt_neox.layers.24.mlp.dense_4h_to_h/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4246">
+				<port id="2" precision="FP32" names="3170,input.99">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6471" name="Constant_278866" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6472" name="Constant_278871" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3968" name="Constant_37482" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2073057360" size="10240" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6473" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3969" name="__module.model.gpt_neox.layers.24.input_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4240">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6474" name="__module.model.gpt_neox.layers.26.attention/aten::cat/Concat" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="3970" name="Constant_37483" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2073067600" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3971" name="__module.model.gpt_neox.layers.24.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4248">
+				<port id="2" precision="FP32" names="3073,hidden_states.97">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6475" name="self.model.gpt_neox.layers.26.attention.rotary_emb.sin_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201162356" size="81920" />
+		<layer id="3972" name="self.model.gpt_neox.layers.24.attention.query_key_value.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 2560" offset="2073077840" size="19660800" />
 			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.26.attention.rotary_emb.sin_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6476" name="self.model.gpt_neox.layers.26.attention.rotary_emb.sin_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="3973" name="Convert_75088" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6477" name="__module.model.gpt_neox.layers.26.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3974" name="self.model.gpt_neox.layers.24.attention.query_key_value.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 1" offset="2092738640" size="7680" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="U8">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6478" name="__module.model.gpt_neox.layers.26.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3975" name="Convert_75091" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>1</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6479" name="__module.model.gpt_neox.layers.26.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3976" name="self.model.gpt_neox.layers.24.attention.query_key_value.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3977" name="self.model.gpt_neox.layers.24.attention.query_key_value.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="7680, 1" offset="2092746320" size="15360" />
+			<output>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I64">
+			</output>
+		</layer>
+		<layer id="3978" name="self.model.gpt_neox.layers.24.attention.query_key_value.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4225,4228,sin.105">
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="2" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6480" name="Constant_38356" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
+		<layer id="3979" name="self.model.gpt_neox.layers.24.attention.query_key_value.weight/fq_weights_1/convert" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I32" />
+				<port id="1" precision="FP32">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6481" name="__module.model.gpt_neox.layers.26.attention/aten::index/Gather_1" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="3980" name="__module.model.gpt_neox.layers.24.attention.query_key_value/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="4233">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6482" name="__module.model.gpt_neox.layers.26.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64" />
-			</input>
+		<layer id="3981" name="Constant_37484" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 7680" offset="2092761680" size="30720" />
 			<output>
-				<port id="2" precision="FP32" names="4234,sin.107">
-					<dim>-1</dim>
+				<port id="0" precision="FP32">
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>1</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6483" name="__module.model.gpt_neox.layers.26.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
+		<layer id="3982" name="__module.model.gpt_neox.layers.24.attention.query_key_value/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>7680</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>1</dim>
+					<dim>7680</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4249">
+				<port id="2" precision="FP32" names="3080,qkv.97">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6484" name="__module.model.gpt_neox.layers.26.attention/aten::add/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="3983" name="__module.model.gpt_neox.layers.24.attention/prim::ListConstruct" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201115992" size="32" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3984" name="__module.model.gpt_neox.layers.24.attention/aten::view/Reshape" type="Reshape" version="opset1">
+			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>7680</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4250">
+				<port id="2" precision="FP32" names="3084,qkv.99">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6485" name="Constant_278917" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3985" name="Constant_15836" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6486" name="Constant_278920" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3986" name="Constant_15838" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6487" name="Constant_278923" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3987" name="Constant_15840" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6488" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3988" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Slice" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4212">
+				<port id="5" precision="FP32" names="3085">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>60</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6489" name="__module.model.gpt_neox.layers.26.attention/aten::cat/Concat_1" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="3989" name="Constant_15849" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
+			<output>
+				<port id="0" precision="I64" names="3086">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="3990" name="__module.model.gpt_neox.layers.24.attention/aten::permute/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4268,4278,query.159,query.161">
+				<port id="2" precision="FP32" names="3087,query.49">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -99290,41 +60593,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6490" name="Constant_278929" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3991" name="Constant_15890" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6491" name="Constant_278932" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3992" name="Constant_15892" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6492" name="Constant_278935" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3993" name="Constant_15894" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6493" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3994" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Slice_1" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -99333,17 +60626,20 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4213,k.53">
+				<port id="5" precision="FP32" names="3094,q.49">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -99351,7 +60647,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6494" name="__module.model.gpt_neox.layers.26.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
+		<layer id="3995" name="__module.model.gpt_neox.layers.24.attention/aten::mul/Multiply" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -99368,7 +60664,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4251">
+				<port id="2" precision="FP32" names="3100">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -99376,160 +60672,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6495" name="Constant_278939" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6496" name="Constant_278938" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3996" name="Constant_16048" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6497" name="__module.model.gpt_neox.layers.26.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6498" name="__module.model.gpt_neox.layers.26.attention/aten::size/Constant_7" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3997" name="Constant_16050" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="6499" name="__module.model.gpt_neox.layers.26.attention/aten::size/Gather_7" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="4252,4257" />
-			</output>
-		</layer>
-		<layer id="6500" name="__module.model.gpt_neox.layers.26.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" />
-			</output>
-		</layer>
-		<layer id="6501" name="__module.model.gpt_neox.layers.26.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-			</input>
-			<output>
-				<port id="1" precision="I64" names="4254,4255,4259,4260" />
-			</output>
-		</layer>
-		<layer id="6502" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="6503" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6504" name="Constant_278937" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="3998" name="Constant_16052" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6505" name="ScatterUpdate_278940" type="ScatterUpdate" version="opset3">
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6506" name="Constant_278943" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6507" name="Constant_278946" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
 			</output>
 		</layer>
-		<layer id="6508" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="3999" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Slice_2" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -99538,50 +60705,31 @@
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4261,x2.107">
+				<port id="5" precision="FP32" names="3110,x2.97">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
 					<dim>10</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="6509" name="Constant_299337_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6510" name="Constant_299337" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="4000" name="Constant_37485" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>1</dim>
@@ -99589,7 +60737,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6511" name="__module.model.gpt_neox.layers.26.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
+		<layer id="4001" name="__module.model.gpt_neox.layers.24.attention/aten::neg/Multiply" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -99606,7 +60754,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4262">
+				<port id="2" precision="FP32" names="3111">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -99614,30 +60762,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6512" name="Constant_278991" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4002" name="Constant_15945" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6513" name="Constant_278996" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4003" name="Constant_15989" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4004" name="Constant_15991" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6514" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4005" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Slice_3" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -99646,17 +60795,20 @@
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4256">
+				<port id="5" precision="FP32" names="3105">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -99664,7 +60816,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6515" name="__module.model.gpt_neox.layers.26.attention/aten::cat/Concat_2" type="Concat" version="opset1">
+		<layer id="4006" name="__module.model.gpt_neox.layers.24.attention/aten::cat/Concat" type="Concat" version="opset1">
 			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
@@ -99681,7 +60833,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4264">
+				<port id="2" precision="FP32" names="3113">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -99689,7 +60841,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6516" name="__module.model.gpt_neox.layers.26.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
+		<layer id="4007" name="__module.model.gpt_neox.layers.24.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -99706,7 +60858,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4265">
+				<port id="2" precision="FP32" names="3114">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -99714,7 +60866,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6517" name="__module.model.gpt_neox.layers.26.attention/aten::add/Add_1" type="Add" version="opset1">
+		<layer id="4008" name="__module.model.gpt_neox.layers.24.attention/aten::add/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -99731,7 +60883,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4266">
+				<port id="2" precision="FP32" names="3115">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -99739,41 +60891,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6518" name="Constant_279041" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4009" name="Constant_15900" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6519" name="Constant_279044" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4010" name="Constant_15902" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6520" name="Constant_279047" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4011" name="Constant_15904" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6521" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4012" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Slice_4" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -99782,17 +60924,20 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4214">
+				<port id="5" precision="FP32" names="3095">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -99800,7 +60945,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6522" name="__module.model.gpt_neox.layers.26.attention/aten::cat/Concat_3" type="Concat" version="opset1">
+		<layer id="4013" name="__module.model.gpt_neox.layers.24.attention/aten::cat/Concat_1" type="Concat" version="opset1">
 			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
@@ -99817,32 +60962,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4270">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6523" name="__module.model.gpt_neox.layers.26.attention/aten::to/Convert_1" type="Concat" version="opset1">
-			<data axis="-2" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="present.26.key">
+				<port id="2" precision="FP32" names="3133">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -99850,49 +60970,37 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6524" name="Constant_338118_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="6525" name="Constant_338118" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
+		<layer id="4014" name="Constant_53127" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="1" precision="FP32" />
+				<port id="0" precision="FP32" />
 			</output>
 		</layer>
-		<layer id="6526" name="Constant_338111" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
+		<layer id="4015" name="Constant_53120" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6527" name="Constant_338113" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
+		<layer id="4016" name="Constant_53122" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6528" name="Constant_338115" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
+		<layer id="4017" name="Constant_53124" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6529" name="Concat_338116" type="Concat" version="opset1">
+		<layer id="4018" name="Concat_53125" type="Concat" version="opset1">
 			<data axis="0" />
 			<input>
 				<port id="0" precision="I64">
@@ -99914,7 +61022,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6530" name="Broadcast_338119" type="Broadcast" version="opset3">
+		<layer id="4019" name="Broadcast_53128" type="Broadcast" version="opset3">
 			<data mode="numpy" />
 			<input>
 				<port id="0" precision="FP32" />
@@ -99931,8 +61039,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6531" name="ReadValue_330503" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.26.valuepresent.26.value" variable_type="f32" variable_shape="?,32,?,80" />
+		<layer id="4020" name="ReadValue_50251" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.24.keypresent.24.key" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -99942,7 +61050,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="past_key_values.26.value">
+				<port id="1" precision="FP32" names="past_key_values.24.key">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -99950,13 +61058,13 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6532" name="Constant_323544" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
+		<layer id="4021" name="Constant_47922" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="6533" name="Gather_323545" type="Gather" version="opset8">
+		<layer id="4022" name="Gather_47923" type="Gather" version="opset8">
 			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="FP32">
@@ -99979,41 +61087,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6534" name="Constant_279053" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4023" name="Constant_15854" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6535" name="Constant_279056" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4024" name="Constant_15856" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6536" name="Constant_279059" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4025" name="Constant_15858" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6537" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4026" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Slice_5" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -100022,17 +61120,20 @@
 					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4208">
+				<port id="5" precision="FP32" names="3088">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -100040,15 +61141,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6538" name="Constant_15357" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="4027" name="Constant_15867" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64" names="4209">
+				<port id="0" precision="I64" names="3089">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6539" name="__module.model.gpt_neox.layers.26.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
+		<layer id="4028" name="__module.model.gpt_neox.layers.24.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -100061,7 +61162,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4210">
+				<port id="2" precision="FP32" names="3090,key.97">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -100069,8 +61170,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6540" name="__module.model.gpt_neox.layers.26.attention/aten::cat/Concat_5" type="Concat" version="opset1">
-			<data axis="-2" />
+		<layer id="4029" name="Constant_15910" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4030" name="Constant_15912" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4031" name="Constant_15914" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4032" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Slice_6" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -100078,176 +61202,303 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="5" precision="FP32" names="3096,k.49">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4033" name="__module.model.gpt_neox.layers.24.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="present.26.value">
+				<port id="2" precision="FP32" names="3116">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6541" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Slice_11" type="Const" version="opset1">
-			<data element_type="boolean" shape="1, 1, 2048, 2048" offset="201244276" size="4194304" />
+		<layer id="4034" name="Constant_16191" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
-				<port id="0" precision="BOOL" names="4287">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6542" name="Constant_279063" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4035" name="Constant_16193" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>3</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6543" name="Constant_279062" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162298" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4036" name="Constant_16195" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6544" name="__module.model.gpt_neox.layers.26.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="4037" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Slice_7" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="5" precision="FP32" names="3126,x2.99">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6545" name="__module.model.gpt_neox.layers.26.attention/aten::size/Constant_8" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4038" name="Constant_37486" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6546" name="__module.model.gpt_neox.layers.26.attention/aten::size/Gather_8" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="4039" name="__module.model.gpt_neox.layers.24.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="4282" />
+				<port id="2" precision="FP32" names="3127">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4040" name="Constant_16088" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6547" name="__module.model.gpt_neox.layers.26.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="4041" name="Constant_16132" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4042" name="Constant_16134" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4043" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Slice_8" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="5" precision="FP32" names="3121">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6548" name="__module.model.gpt_neox.layers.26.attention/aten::size/Constant_9" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4044" name="__module.model.gpt_neox.layers.24.attention/aten::cat/Concat_2" type="Concat" version="opset1">
+			<data axis="-1" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I32" />
+				<port id="2" precision="FP32" names="3129">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6549" name="__module.model.gpt_neox.layers.26.attention/aten::size/Gather_9" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="4045" name="__module.model.gpt_neox.layers.24.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="4280" />
+				<port id="2" precision="FP32" names="3130">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6550" name="__module.model.gpt_neox.layers.26.attention/aten::sub/Subtract" type="Subtract" version="opset1">
+		<layer id="4046" name="__module.model.gpt_neox.layers.24.attention/aten::add/Add_1" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64" names="4284,4285" />
+				<port id="2" precision="FP32" names="3131">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6551" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Constant_4" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4047" name="Constant_15920" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6552" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="4048" name="Constant_15922" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6553" name="Constant_279061" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4049" name="Constant_15924" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6554" name="ScatterUpdate_279064" type="ScatterUpdate" version="opset3">
+		<layer id="4050" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Slice_9" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -100255,280 +61506,313 @@
 				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
 				<port id="4" precision="I64">
-					<dim>3</dim>
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="6555" name="Constant_279065" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="5" precision="FP32" names="3097">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6556" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1">
+		<layer id="4051" name="__module.model.gpt_neox.layers.24.attention/aten::cat/Concat_3" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="3135">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6557" name="ScatterUpdate_279066" type="ScatterUpdate" version="opset3">
+		<layer id="4052" name="__module.model.gpt_neox.layers.24.attention/aten::cat/Concat_4" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="present.24.key">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="3" precision="I32">
+			</output>
+		</layer>
+		<layer id="4053" name="Constant_53139" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
+			<output>
+				<port id="0" precision="FP32" />
+			</output>
+		</layer>
+		<layer id="4054" name="Constant_53132" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="4055" name="Constant_53134" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6558" name="Constant_279069" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438612" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4056" name="Constant_53136" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>3</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6559" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4057" name="Concat_53137" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>3</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>3</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>3</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="BOOL" names="4288">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
+				<port id="4" precision="I64">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6560" name="Constant_279142" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
+		<layer id="4058" name="Broadcast_53140" type="Broadcast" version="opset3">
+			<data mode="numpy" />
+			<input>
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="6561" name="Constant_279143" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6562" name="Constant_279139" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4059" name="ReadValue_50253" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.24.valuepresent.24.value" variable_type="f32" variable_shape="?,32,?,80" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="1" precision="FP32" names="past_key_values.24.value">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6563" name="Constant_279138" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4060" name="Constant_47925" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="6564" name="ScatterUpdate_279144" type="ScatterUpdate" version="opset3">
+		<layer id="4061" name="Gather_47926" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
+				<port id="1" precision="I32">
+					<dim>-1</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="2" precision="I64" />
+			</input>
+			<output>
+				<port id="3" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="3" precision="I32">
+			</output>
+		</layer>
+		<layer id="4062" name="Constant_15872" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="4063" name="Constant_15874" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6565" name="Constant_279147" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4064" name="Constant_15876" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6566" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4065" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Slice_10" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>2048</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="BOOL" names="4289,4290">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="5" precision="FP32" names="3091">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6567" name="__module.model.gpt_neox.layers.26.attention/aten::where/Convert_1_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="6568" name="__module.model.gpt_neox.layers.26.attention/aten::where/Convert_1" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="6569" name="__module.model.gpt_neox.layers.26.attention/aten::full/Convert" type="Const" version="opset1">
-			<data element_type="f32" shape="" offset="205438636" size="4" />
+		<layer id="4066" name="Constant_15885" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="FP32" names="4277" />
+				<port id="0" precision="I64" names="3092">
+					<dim>4</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6570" name="__module.model.gpt_neox.layers.26.attention/aten::where/Select" type="Select" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4067" name="__module.model.gpt_neox.layers.24.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
 			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
-				<port id="1" precision="FP32" />
-				<port id="2" precision="FP32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="4291,causal_mask.105">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="3093">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6571" name="Add_290494" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4068" name="__module.model.gpt_neox.layers.24.attention/aten::cat/Concat_5" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="present.24.value">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4069" name="Constant_16267" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6572" name="ShapeOf_290495" type="ShapeOf" version="opset3">
+		<layer id="4070" name="ShapeOf_40416" type="ShapeOf" version="opset3">
 			<data output_type="i64" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
@@ -100537,27 +61821,21 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6573" name="Constant_254719" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4071" name="Constant_40417" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116168" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6574" name="__module.model.gpt_neox.layers.26.attention/aten::size/Constant_10" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4072" name="Constant_40418" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="6575" name="__module.model.gpt_neox.layers.26.attention/aten::size/Gather_10" type="Gather" version="opset8">
+		<layer id="4073" name="Gather_40419" type="Gather" version="opset8">
 			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="I64">
@@ -100566,16 +61844,16 @@
 				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I32" />
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="4275">
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6576" name="__module.model.gpt_neox.layers.26.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="4074" name="Add_40422" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="I64">
 					<dim>1</dim>
@@ -100583,49 +61861,22 @@
 				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64" names="4292">
-					<dim>4</dim>
+				<port id="2" precision="I64" names="3140">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6577" name="__module.model.gpt_neox.layers.26.attention/aten::expand/Abs" type="Abs" version="opset1">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
+		<layer id="4075" name="Constant_16270" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6578" name="Maximum_290496" type="Maximum" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6579" name="__module.model.gpt_neox.layers.26.attention/aten::add/Add_2" type="Broadcast" version="opset3">
-			<data mode="numpy" />
+		<layer id="4076" name="__module.model.gpt_neox.layers.24.attention/aten::slice/Slice_14" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -100634,19 +61885,28 @@
 					<dim>-1</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4294">
-					<dim>-1</dim>
+				<port id="5" precision="FP32" names="3144">
 					<dim>-1</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
 					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6580" name="__module.model.gpt_neox.layers.26.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+		<layer id="4077" name="__module.model.gpt_neox.layers.24.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
 			<data causal="false" />
 			<input>
 				<port id="0" precision="FP32">
@@ -100669,13 +61929,13 @@
 				</port>
 				<port id="3" precision="FP32">
 					<dim>-1</dim>
-					<dim>-1</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
 					<dim>-1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4295,4296,sdpa_result.53,tensor.105">
+				<port id="4" precision="FP32" names="3145,attn_output.145">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -100683,15 +61943,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6581" name="Constant_15798" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="4078" name="__module.model.gpt_neox.layers.24.attention/aten::transpose/Constant" type="Const" version="opset1">
+			<data element_type="i32" shape="4" offset="201116200" size="16" />
 			<output>
-				<port id="0" precision="I64" names="4297">
+				<port id="0" precision="I32">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6582" name="__module.model.gpt_neox.layers.26.attention/aten::permute/Transpose_3" type="Transpose" version="opset1">
+		<layer id="4079" name="__module.model.gpt_neox.layers.24.attention/aten::transpose/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -100699,12 +61959,12 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="I32">
 					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4298">
+				<port id="2" precision="FP32" names="3146">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -100712,18 +61972,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6583" name="__module.model.gpt_neox.layers.26.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438650" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4080" name="Constant_41008" type="Const" version="opset1">
+			<data element_type="i64" shape="3" offset="201116216" size="24" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>3</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6584" name="__module.model.gpt_neox.layers.26.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
+		<layer id="4081" name="__module.model.gpt_neox.layers.24.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
 			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -100737,15 +61994,15 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4303">
+				<port id="2" precision="FP32" names="3149">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6585" name="self.model.gpt_neox.layers.26.attention.dense.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 2560" offset="2253689554" size="6553600" />
+		<layer id="4082" name="self.model.gpt_neox.layers.24.attention.dense.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 2560" offset="2092792400" size="6553600" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>2560</dim>
@@ -100753,7 +62010,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6586" name="Convert_22474" type="Convert" version="opset1">
+		<layer id="4083" name="Convert_75440" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -100768,8 +62025,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6587" name="self.model.gpt_neox.layers.26.attention.dense.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="2260243154" size="2560" />
+		<layer id="4084" name="self.model.gpt_neox.layers.24.attention.dense.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="2099346000" size="2560" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>2560</dim>
@@ -100777,7 +62034,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6588" name="Convert_22477" type="Convert" version="opset1">
+		<layer id="4085" name="Convert_75443" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -100792,7 +62049,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6589" name="self.model.gpt_neox.layers.26.attention.dense.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="4086" name="self.model.gpt_neox.layers.24.attention.dense.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -100811,8 +62068,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6590" name="self.model.gpt_neox.layers.26.attention.dense.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="2260245714" size="5120" />
+		<layer id="4087" name="self.model.gpt_neox.layers.24.attention.dense.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="2099348560" size="5120" />
 			<output>
 				<port id="0" precision="FP16">
 					<dim>2560</dim>
@@ -100820,7 +62077,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6591" name="self.model.gpt_neox.layers.26.attention.dense.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="4088" name="self.model.gpt_neox.layers.24.attention.dense.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -100839,11 +62096,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6592" name="self.model.gpt_neox.layers.26.attention.dense.weight" type="Convert" version="opset1">
+		<layer id="4089" name="self.model.gpt_neox.layers.24.attention.dense.weight/fq_weights_1/convert" type="Convert" version="opset1">
 			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
 			<input>
 				<port id="0" precision="FP16">
 					<dim>2560</dim>
@@ -100857,7 +62111,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6593" name="__module.model.gpt_neox.layers.26.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="4090" name="__module.model.gpt_neox.layers.24.attention.dense/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -100878,37 +62132,121 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6594" name="Constant_299338_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2260250834" size="5120" />
+		<layer id="4091" name="Constant_37487" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2099353680" size="10240" />
 			<output>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6595" name="Constant_299338" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="4092" name="__module.model.gpt_neox.layers.24.attention.dense/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
+				<port id="2" precision="FP32" names="3152,3154,input.97">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4093" name="__module.model.gpt_neox.layers.24/aten::add/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="3172">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4094" name="__module.model.gpt_neox.layers.24/aten::add/Add_1" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
 				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="3173,3175,hidden_states.103">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4095" name="__module.model.gpt_neox.layers.25.input_layernorm/aten::layer_norm/Multiply" type="Const" version="opset1">
+			<data element_type="i32" shape="1" offset="128867668" size="4" />
+			<output>
+				<port id="0" precision="I32">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4096" name="__module.model.gpt_neox.layers.25.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
+			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="I32">
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4097" name="Constant_37488" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2099363920" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6596" name="__module.model.gpt_neox.layers.26.attention.dense/aten::linear/Add" type="Add" version="opset1">
+		<layer id="4098" name="__module.model.gpt_neox.layers.25.post_attention_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -100923,14 +62261,24 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4306,4308,input.107">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6597" name="__module.model.gpt_neox.layers.26/aten::add/Add" type="Add" version="opset1">
+		<layer id="4099" name="Constant_37489" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2099374160" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4100" name="__module.model.gpt_neox.layers.25.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -100939,59 +62287,338 @@
 					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="3273">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
+			</output>
+		</layer>
+		<layer id="4101" name="self.model.gpt_neox.layers.25.mlp.dense_h_to_4h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 2560" offset="2099384400" size="26214400" />
+			<output>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4102" name="Convert_74684" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4103" name="self.model.gpt_neox.layers.25.mlp.dense_h_to_4h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 1" offset="2125598800" size="10240" />
+			<output>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4104" name="Convert_74687" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4105" name="self.model.gpt_neox.layers.25.mlp.dense_h_to_4h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4106" name="self.model.gpt_neox.layers.25.mlp.dense_h_to_4h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="10240, 1" offset="2125609040" size="20480" />
+			<output>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4107" name="self.model.gpt_neox.layers.25.mlp.dense_h_to_4h.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4108" name="__module.model.gpt_neox.layers.25.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<rt_info>
+				<attribute name="decompression" version="0" />
+			</rt_info>
+			<input>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4326">
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4109" name="__module.model.gpt_neox.layers.25.mlp.dense_h_to_4h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
+			<input>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4110" name="self.model.gpt_neox.layers.25.mlp.dense_h_to_4h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="10240" offset="2125629520" size="20480" />
+			<output>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.25.mlp.dense_h_to_4h.bias">
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4111" name="__module.model.gpt_neox.layers.25.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<rt_info>
+				<attribute name="decompression" version="0" />
+			</rt_info>
+			<input>
+				<port id="0" precision="BF16">
+					<dim>10240</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6598" name="__module.model.gpt_neox.layers.26/aten::add/Add_1" type="Add" version="opset1">
+		<layer id="4112" name="__module.model.gpt_neox.layers.25.mlp.dense_h_to_4h/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 				<port id="1" precision="FP32">
+					<dim>10240</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="3278">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4113" name="__module.model.gpt_neox.layers.25.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
+			<data approximation_mode="ERF" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4327,4329,hidden_states.55">
+				<port id="1" precision="FP32" names="3279">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4114" name="self.model.gpt_neox.layers.25.mlp.dense_4h_to_h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 10240" offset="2125650000" size="26214400" />
+			<output>
+				<port id="0" precision="U8">
 					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6599" name="__module.model.gpt_neox.layers.27.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867670" size="4" />
+		<layer id="4115" name="Convert_74694" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I32">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4116" name="self.model.gpt_neox.layers.25.mlp.dense_4h_to_h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="2151864400" size="2560" />
+			<output>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6600" name="__module.model.gpt_neox.layers.27.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
-			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
+		<layer id="4117" name="Convert_74697" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4118" name="self.model.gpt_neox.layers.25.mlp.dense_4h_to_h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4119" name="self.model.gpt_neox.layers.25.mlp.dense_4h_to_h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="2151866960" size="5120" />
+			<output>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4120" name="self.model.gpt_neox.layers.25.mlp.dense_4h_to_h.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4121" name="__module.model.gpt_neox.layers.25.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<rt_info>
+				<attribute name="decompression" version="0" />
+			</rt_info>
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4122" name="__module.model.gpt_neox.layers.25.mlp.dense_4h_to_h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="1" precision="I32">
-					<dim>1</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
@@ -101002,37 +62629,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6601" name="Constant_299339_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2260255954" size="5120" />
+		<layer id="4123" name="self.model.gpt_neox.layers.25.mlp.dense_4h_to_h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="2560" offset="2151872080" size="5120" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.25.mlp.dense_4h_to_h.bias">
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6602" name="Constant_299339" type="Convert" version="opset1">
+		<layer id="4124" name="__module.model.gpt_neox.layers.25.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
 			</rt_info>
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16">
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6603" name="__module.model.gpt_neox.layers.27.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
+		<layer id="4125" name="__module.model.gpt_neox.layers.25.mlp.dense_4h_to_h/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -101041,50 +62662,60 @@
 					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="3282,input.103">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6604" name="Constant_299340_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2260261074" size="5120" />
+		<layer id="4126" name="Constant_37490" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2151877200" size="10240" />
 			<output>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6605" name="Constant_299340" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="4127" name="__module.model.gpt_neox.layers.25.input_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4128" name="Constant_37491" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2151887440" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6606" name="__module.model.gpt_neox.layers.27.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+		<layer id="4129" name="__module.model.gpt_neox.layers.25.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -101099,127 +62730,124 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4464">
+				<port id="2" precision="FP32" names="3185,hidden_states.101">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6607" name="self.model.gpt_neox.layers.27.mlp.dense_h_to_4h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 2560" offset="2260266194" size="26214400" />
+		<layer id="4130" name="self.model.gpt_neox.layers.25.attention.query_key_value.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 2560" offset="2151897680" size="19660800" />
 			<output>
 				<port id="0" precision="U8">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6608" name="Convert_22524" type="Convert" version="opset1">
+		<layer id="4131" name="Convert_75099" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP16">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6609" name="self.model.gpt_neox.layers.27.mlp.dense_h_to_4h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 1" offset="2286480594" size="10240" />
+		<layer id="4132" name="self.model.gpt_neox.layers.25.attention.query_key_value.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 1" offset="2171558480" size="7680" />
 			<output>
 				<port id="0" precision="U8">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6610" name="Convert_22527" type="Convert" version="opset1">
+		<layer id="4133" name="Convert_75102" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP16">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6611" name="self.model.gpt_neox.layers.27.mlp.dense_h_to_4h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="4134" name="self.model.gpt_neox.layers.25.attention.query_key_value.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP16">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
 				<port id="2" precision="FP16">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6612" name="self.model.gpt_neox.layers.27.mlp.dense_h_to_4h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="10240, 1" offset="2286490834" size="20480" />
+		<layer id="4135" name="self.model.gpt_neox.layers.25.attention.query_key_value.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="7680, 1" offset="2171566160" size="15360" />
 			<output>
 				<port id="0" precision="FP16">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6613" name="self.model.gpt_neox.layers.27.mlp.dense_h_to_4h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="4136" name="self.model.gpt_neox.layers.25.attention.query_key_value.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP16">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
 				<port id="2" precision="FP16">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6614" name="self.model.gpt_neox.layers.27.mlp.dense_h_to_4h.weight" type="Convert" version="opset1">
+		<layer id="4137" name="self.model.gpt_neox.layers.25.attention.query_key_value.weight/fq_weights_1/convert" type="Convert" version="opset1">
 			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
 			<input>
 				<port id="0" precision="FP16">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP32">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6615" name="__module.model.gpt_neox.layers.27.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="4138" name="__module.model.gpt_neox.layers.25.attention.query_key_value/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -101228,7 +62856,7 @@
 					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>10240</dim>
+					<dim>7680</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
@@ -101236,859 +62864,1101 @@
 				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6616" name="Constant_299341_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 10240" offset="2286511314" size="20480" />
+		<layer id="4139" name="Constant_37492" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 7680" offset="2171581520" size="30720" />
 			<output>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>10240</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6617" name="Constant_299341" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="4140" name="__module.model.gpt_neox.layers.25.attention.query_key_value/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>7680</dim>
+				</port>
+				<port id="1" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>10240</dim>
+					<dim>7680</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
+				<port id="2" precision="FP32" names="3192,qkv.101">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>7680</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4141" name="__module.model.gpt_neox.layers.25.attention/prim::ListConstruct" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201115992" size="32" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4142" name="__module.model.gpt_neox.layers.25.attention/aten::view/Reshape" type="Reshape" version="opset1">
+			<data special_zero="true" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>7680</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>4</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="3196,qkv.103">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4143" name="Constant_16466" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4144" name="Constant_16468" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6618" name="__module.model.gpt_neox.layers.27.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4145" name="Constant_16470" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4146" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Slice" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
-				<port id="1" precision="FP32">
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4469">
+				<port id="5" precision="FP32" names="3197">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6619" name="__module.model.gpt_neox.layers.27.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
-			<data approximation_mode="ERF" />
+		<layer id="4147" name="Constant_16479" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
+			<output>
+				<port id="0" precision="I64" names="3198">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4148" name="__module.model.gpt_neox.layers.25.attention/aten::permute/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="4470">
+				<port id="2" precision="FP32" names="3199,query.51">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6620" name="self.model.gpt_neox.layers.27.mlp.dense_4h_to_h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 10240" offset="2286531794" size="26214400" />
+		<layer id="4149" name="Constant_16520" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6621" name="Convert_22534" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
+		<layer id="4150" name="Constant_16522" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6622" name="self.model.gpt_neox.layers.27.mlp.dense_4h_to_h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="2312746194" size="2560" />
+		<layer id="4151" name="Constant_16524" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6623" name="Convert_22537" type="Convert" version="opset1">
-			<data destination_type="f16" />
+		<layer id="4152" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Slice_1" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
+				<port id="5" precision="FP32" names="3206,q.51">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6624" name="self.model.gpt_neox.layers.27.mlp.dense_4h_to_h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="4153" name="__module.model.gpt_neox.layers.25.attention/aten::mul/Multiply" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
 					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="2" precision="FP32" names="3212">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6625" name="self.model.gpt_neox.layers.27.mlp.dense_4h_to_h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="2312748754" size="5120" />
+		<layer id="4154" name="Constant_16678" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6626" name="self.model.gpt_neox.layers.27.mlp.dense_4h_to_h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+		<layer id="4155" name="Constant_16680" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="4156" name="Constant_16682" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6627" name="self.model.gpt_neox.layers.27.mlp.dense_4h_to_h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="4157" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Slice_2" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="5" precision="FP32" names="3222,x2.101">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6628" name="__module.model.gpt_neox.layers.27.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
+		<layer id="4158" name="Constant_37493" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4159" name="__module.model.gpt_neox.layers.25.attention/aten::neg/Multiply" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>10</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="3223">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6629" name="Constant_299342_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2312753874" size="5120" />
+		<layer id="4160" name="Constant_16575" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="FP16">
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4161" name="Constant_16619" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6630" name="Constant_299342" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="4162" name="Constant_16621" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4163" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Slice_3" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
+				<port id="5" precision="FP32" names="3217">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4164" name="__module.model.gpt_neox.layers.25.attention/aten::cat/Concat" type="Concat" version="opset1">
+			<data axis="-1" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="3225">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6631" name="__module.model.gpt_neox.layers.27.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1">
+		<layer id="4165" name="__module.model.gpt_neox.layers.25.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
+					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4473,input.113">
+				<port id="2" precision="FP32" names="3226">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6632" name="Constant_299343_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2312758994" size="5120" />
+		<layer id="4166" name="__module.model.gpt_neox.layers.25.attention/aten::add/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="FP16">
+				<port id="2" precision="FP32" names="3227">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4167" name="Constant_16530" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4168" name="Constant_16532" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6633" name="Constant_299343" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="4169" name="Constant_16534" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4170" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Slice_4" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+				<port id="5" precision="FP32" names="3207">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6634" name="__module.model.gpt_neox.layers.27.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4171" name="__module.model.gpt_neox.layers.25.attention/aten::cat/Concat_1" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="3245">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4172" name="Constant_53151" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
+			<output>
+				<port id="0" precision="FP32" />
+			</output>
+		</layer>
+		<layer id="4173" name="Constant_53144" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4174" name="Constant_53146" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4175" name="Constant_53148" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4176" name="Concat_53149" type="Concat" version="opset1">
+			<data axis="0" />
+			<input>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6635" name="Constant_299344_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2312764114" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+				<port id="4" precision="I64">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6636" name="Constant_299344" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="4177" name="Broadcast_53152" type="Broadcast" version="opset3">
+			<data mode="numpy" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6637" name="__module.model.gpt_neox.layers.27.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4178" name="ReadValue_50255" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.25.keypresent.25.key" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4339">
+				<port id="1" precision="FP32" names="past_key_values.25.key">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6638" name="self.model.gpt_neox.layers.27.attention.query_key_value.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 2560" offset="2312769234" size="19660800" />
+		<layer id="4179" name="Constant_47928" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="6639" name="Convert_22504" type="Convert" version="opset1">
-			<data destination_type="f16" />
+		<layer id="4180" name="Gather_47929" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I32">
+					<dim>-1</dim>
 				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="3" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6640" name="self.model.gpt_neox.layers.27.attention.query_key_value.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 1" offset="2332430034" size="7680" />
+		<layer id="4181" name="Constant_16484" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6641" name="Convert_22507" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
+		<layer id="4182" name="Constant_16486" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="4183" name="Constant_16488" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6642" name="self.model.gpt_neox.layers.27.attention.query_key_value.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4184" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Slice_5" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
+				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6643" name="self.model.gpt_neox.layers.27.attention.query_key_value.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="7680, 1" offset="2332437714" size="15360" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
+				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="6644" name="self.model.gpt_neox.layers.27.attention.query_key_value.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="3" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="5" precision="FP32" names="3200">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6645" name="self.model.gpt_neox.layers.27.attention.query_key_value.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="4185" name="Constant_16497" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64" names="3201">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6646" name="__module.model.gpt_neox.layers.27.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
+		<layer id="4186" name="__module.model.gpt_neox.layers.25.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="3202,key.101">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6647" name="Constant_299345_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 7680" offset="2332453074" size="15360" />
+		<layer id="4187" name="Constant_16540" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6648" name="Constant_299345" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+		<layer id="4188" name="Constant_16542" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>7680</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="4189" name="Constant_16544" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6649" name="__module.model.gpt_neox.layers.27.attention.query_key_value/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4190" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Slice_6" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>7680</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4346,qkv.109">
+				<port id="5" precision="FP32" names="3208,k.51">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6650" name="__module.model.gpt_neox.layers.27.attention/prim::ListConstruct/Concat" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080154" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6651" name="__module.model.gpt_neox.layers.27.attention/aten::view/Reshape" type="Reshape" version="opset1">
-			<data special_zero="true" />
+		<layer id="4191" name="__module.model.gpt_neox.layers.25.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4350,qkv.111">
-					<dim>-1</dim>
+				<port id="2" precision="FP32" names="3228">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>240</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6652" name="Constant_279154" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4192" name="Constant_16821" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6653" name="Constant_279157" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4193" name="Constant_16823" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6654" name="Constant_279160" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4194" name="Constant_16825" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6655" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4195" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Slice_7" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>240</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4351">
-					<dim>-1</dim>
+				<port id="5" precision="FP32" names="3238,x2.103">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6656" name="Constant_15910" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="4196" name="Constant_37494" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="0" precision="I64" names="4352">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6657" name="__module.model.gpt_neox.layers.27.attention/aten::permute/Transpose" type="Transpose" version="opset1">
+		<layer id="4197" name="__module.model.gpt_neox.layers.25.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4353,query.163">
+				<port id="2" precision="FP32" names="3239">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6658" name="Constant_279166" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4198" name="Constant_16718" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6659" name="Constant_279169" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4199" name="Constant_16762" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6660" name="Constant_279172" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4200" name="Constant_16764" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6661" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4201" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Slice_8" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4360,q.55">
+				<port id="5" precision="FP32" names="3233">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6662" name="self.model.gpt_neox.layers.27.attention.rotary_emb.cos_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201080346" size="81920" />
+		<layer id="4202" name="__module.model.gpt_neox.layers.25.attention/aten::cat/Concat_2" type="Concat" version="opset1">
+			<data axis="-1" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.27.attention.rotary_emb.cos_cached">
-					<dim>2048</dim>
+				<port id="2" precision="FP32" names="3241">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6663" name="self.model.gpt_neox.layers.27.attention.rotary_emb.cos_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="4203" name="__module.model.gpt_neox.layers.25.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
+				<port id="2" precision="FP32" names="3242">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6664" name="__module.model.gpt_neox.layers.27.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4204" name="__module.model.gpt_neox.layers.25.attention/aten::add/Add_1" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="3243">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6665" name="Constant_279178" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4205" name="Constant_16550" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6666" name="Constant_279181" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4206" name="Constant_16552" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6667" name="Constant_279184" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4207" name="Constant_16554" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6668" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4208" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Slice_9" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>240</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4354">
-					<dim>-1</dim>
+				<port id="5" precision="FP32" names="3209">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6669" name="Constant_15924" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="4355">
-					<dim>4</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6670" name="__module.model.gpt_neox.layers.27.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
+		<layer id="4209" name="__module.model.gpt_neox.layers.25.attention/aten::cat/Concat_3" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4356,key.163">
+				<port id="2" precision="FP32" names="3247">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -102096,8 +63966,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6671" name="__module.model.gpt_neox.layers.27.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="4210" name="__module.model.gpt_neox.layers.25.attention/aten::cat/Concat_4" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -102105,78 +63975,53 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6672" name="__module.model.gpt_neox.layers.27.attention/aten::size/Constant_2" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="6673" name="__module.model.gpt_neox.layers.27.attention/aten::size/Gather_2" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="4364" />
-			</output>
-		</layer>
-		<layer id="6674" name="Constant_338131_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
+				<port id="2" precision="FP32" names="present.25.key">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6675" name="Constant_338131" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
+		<layer id="4211" name="Constant_53163" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="1" precision="FP32" />
+				<port id="0" precision="FP32" />
 			</output>
 		</layer>
-		<layer id="6676" name="Constant_338124" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
+		<layer id="4212" name="Constant_53156" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6677" name="Constant_338126" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
+		<layer id="4213" name="Constant_53158" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6678" name="Constant_338128" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
+		<layer id="4214" name="Constant_53160" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6679" name="Concat_338129" type="Concat" version="opset1">
+		<layer id="4215" name="Concat_53161" type="Concat" version="opset1">
 			<data axis="0" />
 			<input>
 				<port id="0" precision="I64">
@@ -102198,7 +64043,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6680" name="Broadcast_338132" type="Broadcast" version="opset3">
+		<layer id="4216" name="Broadcast_53164" type="Broadcast" version="opset3">
 			<data mode="numpy" />
 			<input>
 				<port id="0" precision="FP32" />
@@ -102215,8 +64060,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6681" name="ReadValue_330505" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.27.keypresent.27.key" variable_type="f32" variable_shape="?,32,?,80" />
+		<layer id="4217" name="ReadValue_50257" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.25.valuepresent.25.value" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -102226,7 +64071,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="past_key_values.27.key">
+				<port id="1" precision="FP32" names="past_key_values.25.value">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -102234,120 +64079,66 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6682" name="Constant_323547" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
+		<layer id="4218" name="Constant_47931" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="6683" name="Gather_323548" type="Gather" version="opset8">
+		<layer id="4219" name="Gather_47932" type="Gather" version="opset8">
 			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I64" />
-			</input>
-			<output>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6684" name="__module.model.gpt_neox.layers.27.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6685" name="__module.model.gpt_neox.layers.27.attention/aten::size/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="6686" name="__module.model.gpt_neox.layers.27.attention/aten::size/Gather_3" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="4366" />
-			</output>
-		</layer>
-		<layer id="6687" name="__module.model.gpt_neox.layers.27.attention/aten::add_/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I32">
+					<dim>-1</dim>
+				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="I64" names="4365,4369,4370,seq_len.109" />
+				<port id="3" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6688" name="__module.model.gpt_neox.layers.27.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4220" name="Constant_16502" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6689" name="__module.model.gpt_neox.layers.27.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="4221" name="Constant_16504" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6690" name="__module.model.gpt_neox.layers.27.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4222" name="Constant_16506" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6691" name="__module.model.gpt_neox.layers.27.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4223" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Slice_10" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -102358,114 +64149,89 @@
 				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4373,4376,cos.109">
+				<port id="5" precision="FP32" names="3203">
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6692" name="Constant_38359" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="6693" name="__module.model.gpt_neox.layers.27.attention/aten::index/Gather" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
+		<layer id="4224" name="Constant_16515" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="3" precision="FP32" names="4379">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="I64" names="3204">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6694" name="__module.model.gpt_neox.layers.27.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1">
+		<layer id="4225" name="__module.model.gpt_neox.layers.25.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
-				<port id="1" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4380,cos.111">
+				<port id="2" precision="FP32" names="3205">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6695" name="__module.model.gpt_neox.layers.27.attention/aten::mul/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4226" name="__module.model.gpt_neox.layers.25.attention/aten::cat/Concat_5" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4384">
+				<port id="2" precision="FP32" names="present.25.value">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6696" name="Constant_279189" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6697" name="Constant_279188" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4227" name="Constant_16897" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6698" name="__module.model.gpt_neox.layers.27.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3">
+		<layer id="4228" name="ShapeOf_40487" type="ShapeOf" version="opset3">
 			<data output_type="i64" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
@@ -102474,1259 +64240,1139 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6699" name="__module.model.gpt_neox.layers.27.attention/aten::size/Constant_5" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4229" name="Constant_40488" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116168" size="8" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="6700" name="__module.model.gpt_neox.layers.27.attention/aten::size/Gather_5" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="4385,4390" />
 			</output>
 		</layer>
-		<layer id="6701" name="__module.model.gpt_neox.layers.27.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
+		<layer id="4230" name="Constant_40489" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="2" precision="I64" />
-			</output>
-		</layer>
-		<layer id="6702" name="__module.model.gpt_neox.layers.27.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1">
-			<input>
 				<port id="0" precision="I64" />
-			</input>
-			<output>
-				<port id="1" precision="I64" names="4387,4388,4392,4393" />
-			</output>
-		</layer>
-		<layer id="6703" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Constant_1" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
 			</output>
 		</layer>
-		<layer id="6704" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1">
+		<layer id="4231" name="Gather_40490" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="I64">
+					<dim>4</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="6705" name="Constant_279187" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+				<port id="2" precision="I64" />
+			</input>
 			<output>
-				<port id="0" precision="I32">
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6706" name="ScatterUpdate_279190" type="ScatterUpdate" version="opset3">
+		<layer id="4232" name="Add_40493" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6707" name="Constant_279193" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="I64" names="3252">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6708" name="Constant_279196" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4233" name="Constant_16900" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6709" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4234" name="__module.model.gpt_neox.layers.25.attention/aten::slice/Slice_14" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="4394,x2.109">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6710" name="Constant_299346_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="6711" name="Constant_299346" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="5" precision="FP32" names="3256">
+					<dim>-1</dim>
 					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6712" name="__module.model.gpt_neox.layers.27.attention/aten::neg/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4235" name="__module.model.gpt_neox.layers.25.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+			<data causal="false" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="3" precision="FP32">
+					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4395">
+				<port id="4" precision="FP32" names="3257,attn_output.151">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6713" name="Constant_279241" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6714" name="Constant_279246" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4236" name="__module.model.gpt_neox.layers.25.attention/aten::transpose/Constant" type="Const" version="opset1">
+			<data element_type="i32" shape="4" offset="201116200" size="16" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="I32">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6715" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4237" name="__module.model.gpt_neox.layers.25.attention/aten::transpose/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>80</dim>
 				</port>
-				<port id="3" precision="I64">
+				<port id="1" precision="I32">
 					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4389">
+				<port id="2" precision="FP32" names="3258">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6716" name="__module.model.gpt_neox.layers.27.attention/aten::cat/Concat" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="4238" name="Constant_41009" type="Const" version="opset1">
+			<data element_type="i64" shape="3" offset="201116216" size="24" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>3</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4239" name="__module.model.gpt_neox.layers.25.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
+			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>3</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4397">
+				<port id="2" precision="FP32" names="3261">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6717" name="self.model.gpt_neox.layers.27.attention.rotary_emb.sin_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201162356" size="81920" />
+		<layer id="4240" name="self.model.gpt_neox.layers.25.attention.dense.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 2560" offset="2171612240" size="6553600" />
 			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.27.attention.rotary_emb.sin_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6718" name="self.model.gpt_neox.layers.27.attention.rotary_emb.sin_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="4241" name="Convert_75451" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6719" name="__module.model.gpt_neox.layers.27.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4242" name="self.model.gpt_neox.layers.25.attention.dense.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="2178165840" size="2560" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="U8">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6720" name="__module.model.gpt_neox.layers.27.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4243" name="Convert_75454" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6721" name="__module.model.gpt_neox.layers.27.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4244" name="self.model.gpt_neox.layers.25.attention.dense.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4374,4377,sin.109">
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6722" name="Constant_38362" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
+		<layer id="4245" name="self.model.gpt_neox.layers.25.attention.dense.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="2178168400" size="5120" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6723" name="__module.model.gpt_neox.layers.27.attention/aten::index/Gather_1" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="4246" name="self.model.gpt_neox.layers.25.attention.dense.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
 				</port>
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="4382">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6724" name="__module.model.gpt_neox.layers.27.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1">
+		<layer id="4247" name="self.model.gpt_neox.layers.25.attention.dense.weight/fq_weights_1/convert" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4383,sin.111">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6725" name="__module.model.gpt_neox.layers.27.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4248" name="__module.model.gpt_neox.layers.25.attention.dense/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4398">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4249" name="Constant_37495" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2178173520" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6726" name="__module.model.gpt_neox.layers.27.attention/aten::add/Add" type="Add" version="opset1">
+		<layer id="4250" name="__module.model.gpt_neox.layers.25.attention.dense/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4399">
+				<port id="2" precision="FP32" names="3264,3266,input.101">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6727" name="Constant_279292" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6728" name="Constant_279295" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6729" name="Constant_279298" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6730" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4251" name="__module.model.gpt_neox.layers.25/aten::add/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4361">
+				<port id="2" precision="FP32" names="3284">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>60</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6731" name="__module.model.gpt_neox.layers.27.attention/aten::cat/Concat_1" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="4252" name="__module.model.gpt_neox.layers.25/aten::add/Add_1" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>60</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4417,4427,query.165,query.167">
+				<port id="2" precision="FP32" names="3285,3287,hidden_states.107">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6732" name="Constant_279304" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6733" name="Constant_279307" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6734" name="Constant_279310" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4253" name="__module.model.gpt_neox.layers.26.input_layernorm/aten::layer_norm/Multiply" type="Const" version="opset1">
+			<data element_type="i32" shape="1" offset="128867668" size="4" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="I32">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6735" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4254" name="__module.model.gpt_neox.layers.26.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
+			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="I32">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4362,k.55">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4255" name="Constant_37496" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2178183760" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6736" name="__module.model.gpt_neox.layers.27.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
+		<layer id="4256" name="__module.model.gpt_neox.layers.26.post_attention_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4400">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6737" name="Constant_279314" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6738" name="Constant_279313" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4257" name="Constant_37497" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2178194000" size="10240" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="FP32">
+					<dim>1</dim>
 					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6739" name="__module.model.gpt_neox.layers.27.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="4258" name="__module.model.gpt_neox.layers.26.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="3385">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6740" name="__module.model.gpt_neox.layers.27.attention/aten::size/Constant_7" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4259" name="self.model.gpt_neox.layers.26.mlp.dense_h_to_4h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 2560" offset="2178204240" size="26214400" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6741" name="__module.model.gpt_neox.layers.27.attention/aten::size/Gather_7" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="4260" name="Convert_74704" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="4401,4406" />
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6742" name="__module.model.gpt_neox.layers.27.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
+		<layer id="4261" name="self.model.gpt_neox.layers.26.mlp.dense_h_to_4h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 1" offset="2204418640" size="10240" />
 			<output>
-				<port id="2" precision="I64" />
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6743" name="__module.model.gpt_neox.layers.27.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1">
+		<layer id="4262" name="Convert_74707" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64" />
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64" names="4403,4404,4408,4409" />
-			</output>
-		</layer>
-		<layer id="6744" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6745" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1">
+		<layer id="4263" name="self.model.gpt_neox.layers.26.mlp.dense_h_to_4h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6746" name="Constant_279312" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4264" name="self.model.gpt_neox.layers.26.mlp.dense_h_to_4h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="10240, 1" offset="2204428880" size="20480" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6747" name="ScatterUpdate_279315" type="ScatterUpdate" version="opset3">
+		<layer id="4265" name="self.model.gpt_neox.layers.26.mlp.dense_h_to_4h.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6748" name="Constant_279318" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
+		<layer id="4266" name="__module.model.gpt_neox.layers.26.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<rt_info>
-				<attribute name="precise" version="0" />
+				<attribute name="decompression" version="0" />
 			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+			<input>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="6749" name="Constant_279321" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6750" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4267" name="__module.model.gpt_neox.layers.26.mlp.dense_h_to_4h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4410,x2.111">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6751" name="Constant_299347_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
+		<layer id="4268" name="self.model.gpt_neox.layers.26.mlp.dense_h_to_4h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="10240" offset="2204449360" size="20480" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.26.mlp.dense_h_to_4h.bias">
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6752" name="Constant_299347" type="Convert" version="opset1">
+		<layer id="4269" name="__module.model.gpt_neox.layers.26.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
 			</rt_info>
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16">
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6753" name="__module.model.gpt_neox.layers.27.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
+		<layer id="4270" name="__module.model.gpt_neox.layers.26.mlp.dense_h_to_4h/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>10240</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4411">
+				<port id="2" precision="FP32" names="3390">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6754" name="Constant_279366" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4271" name="__module.model.gpt_neox.layers.26.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
+			<data approximation_mode="ERF" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32" names="3391">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6755" name="Constant_279371" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4272" name="self.model.gpt_neox.layers.26.mlp.dense_4h_to_h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 10240" offset="2204469840" size="26214400" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6756" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4273" name="Convert_74714" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+			</input>
+			<output>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+			</output>
+		</layer>
+		<layer id="4274" name="self.model.gpt_neox.layers.26.mlp.dense_4h_to_h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="2230684240" size="2560" />
+			<output>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+			</output>
+		</layer>
+		<layer id="4275" name="Convert_74717" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4405">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6757" name="__module.model.gpt_neox.layers.27.attention/aten::cat/Concat_2" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="4276" name="self.model.gpt_neox.layers.26.mlp.dense_4h_to_h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4413">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4277" name="self.model.gpt_neox.layers.26.mlp.dense_4h_to_h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="2230686800" size="5120" />
+			<output>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6758" name="__module.model.gpt_neox.layers.27.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
+		<layer id="4278" name="self.model.gpt_neox.layers.26.mlp.dense_4h_to_h.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4414">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6759" name="__module.model.gpt_neox.layers.27.attention/aten::add/Add_1" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4279" name="__module.model.gpt_neox.layers.26.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<rt_info>
+				<attribute name="decompression" version="0" />
+			</rt_info>
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4280" name="__module.model.gpt_neox.layers.26.mlp.dense_4h_to_h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>10240</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4415">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6760" name="Constant_279416" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4281" name="self.model.gpt_neox.layers.26.mlp.dense_4h_to_h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="2560" offset="2230691920" size="5120" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.26.mlp.dense_4h_to_h.bias">
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6761" name="Constant_279419" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
+		<layer id="4282" name="__module.model.gpt_neox.layers.26.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<rt_info>
-				<attribute name="precise" version="0" />
+				<attribute name="decompression" version="0" />
 			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+			<input>
+				<port id="0" precision="BF16">
+					<dim>2560</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="6762" name="Constant_279422" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6763" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4283" name="__module.model.gpt_neox.layers.26.mlp.dense_4h_to_h/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4363">
+				<port id="2" precision="FP32" names="3394,input.107">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>60</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6764" name="__module.model.gpt_neox.layers.27.attention/aten::cat/Concat_3" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="4284" name="Constant_37498" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2230697040" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4285" name="__module.model.gpt_neox.layers.26.input_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4419">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6765" name="__module.model.gpt_neox.layers.27.attention/aten::to/Convert_1" type="Concat" version="opset1">
-			<data axis="-2" />
+		<layer id="4286" name="Constant_37499" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2230707280" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4287" name="__module.model.gpt_neox.layers.26.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="present.27.key">
+				<port id="2" precision="FP32" names="3297,hidden_states.105">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6766" name="Constant_338144_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
+		<layer id="4288" name="self.model.gpt_neox.layers.26.attention.query_key_value.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 2560" offset="2230717520" size="19660800" />
 			<output>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6767" name="Constant_338144" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="4289" name="Convert_75110" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" />
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6768" name="Constant_338137" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
+		<layer id="4290" name="self.model.gpt_neox.layers.26.attention.query_key_value.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 1" offset="2250378320" size="7680" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="U8">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6769" name="Constant_338139" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
+		<layer id="4291" name="Convert_75113" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>1</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6770" name="Constant_338141" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
+		<layer id="4292" name="self.model.gpt_neox.layers.26.attention.query_key_value.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
+					<dim>1</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
+				<port id="2" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4293" name="self.model.gpt_neox.layers.26.attention.query_key_value.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="7680, 1" offset="2250386000" size="15360" />
+			<output>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6771" name="Concat_338142" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="4294" name="self.model.gpt_neox.layers.26.attention.query_key_value.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
+			</output>
+		</layer>
+		<layer id="4295" name="self.model.gpt_neox.layers.26.attention.query_key_value.weight/fq_weights_1/convert" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6772" name="Broadcast_338145" type="Broadcast" version="opset3">
-			<data mode="numpy" />
+		<layer id="4296" name="__module.model.gpt_neox.layers.26.attention.query_key_value/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6773" name="ReadValue_330507" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.27.valuepresent.27.value" variable_type="f32" variable_shape="?,32,?,80" />
+		<layer id="4297" name="Constant_37500" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 7680" offset="2250401360" size="30720" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>7680</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4298" name="__module.model.gpt_neox.layers.26.attention.query_key_value/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>7680</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>7680</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="past_key_values.27.value">
+				<port id="2" precision="FP32" names="3304,qkv.105">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6774" name="Constant_323550" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
+		<layer id="4299" name="__module.model.gpt_neox.layers.26.attention/prim::ListConstruct" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201115992" size="32" />
 			<output>
-				<port id="0" precision="I64" />
+				<port id="0" precision="I64">
+					<dim>4</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6775" name="Gather_323551" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="4300" name="__module.model.gpt_neox.layers.26.attention/aten::view/Reshape" type="Reshape" version="opset1">
+			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>7680</dim>
 				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
-				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="3" precision="FP32">
+				<port id="2" precision="FP32" names="3308,qkv.107">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6776" name="Constant_279428" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4301" name="Constant_17096" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6777" name="Constant_279431" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4302" name="Constant_17098" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6778" name="Constant_279434" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4303" name="Constant_17100" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6779" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4304" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Slice" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -103735,17 +65381,20 @@
 					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4357">
+				<port id="5" precision="FP32" names="3309">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -103753,15 +65402,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6780" name="Constant_15938" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="4305" name="Constant_17109" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64" names="4358">
+				<port id="0" precision="I64" names="3310">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6781" name="__module.model.gpt_neox.layers.27.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
+		<layer id="4306" name="__module.model.gpt_neox.layers.26.attention/aten::permute/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -103774,32 +65423,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4359">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6782" name="__module.model.gpt_neox.layers.27.attention/aten::cat/Concat_5" type="Concat" version="opset1">
-			<data axis="-2" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="present.27.value">
+				<port id="2" precision="FP32" names="3311,query.53">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -103807,41 +65431,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6783" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Slice_11" type="Const" version="opset1">
-			<data element_type="boolean" shape="1, 1, 2048, 2048" offset="201244276" size="4194304" />
+		<layer id="4307" name="Constant_17150" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="BOOL" names="4436">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6784" name="Constant_279438" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4308" name="Constant_17152" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>3</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6785" name="Constant_279437" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162298" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4309" name="Constant_17154" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6786" name="__module.model.gpt_neox.layers.27.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="4310" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Slice_1" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -103849,118 +65463,84 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-			</input>
-			<output>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="6787" name="__module.model.gpt_neox.layers.27.attention/aten::size/Constant_8" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="6788" name="__module.model.gpt_neox.layers.27.attention/aten::size/Gather_8" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="4431" />
+				<port id="5" precision="FP32" names="3318,q.53">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6789" name="__module.model.gpt_neox.layers.27.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="4311" name="__module.model.gpt_neox.layers.26.attention/aten::mul/Multiply" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="3324">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6790" name="__module.model.gpt_neox.layers.27.attention/aten::size/Constant_9" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4312" name="Constant_17308" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="6791" name="__module.model.gpt_neox.layers.27.attention/aten::size/Gather_9" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="4429" />
-			</output>
-		</layer>
-		<layer id="6792" name="__module.model.gpt_neox.layers.27.attention/aten::sub/Subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" names="4433,4434" />
-			</output>
-		</layer>
-		<layer id="6793" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Constant_4" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
 			</output>
 		</layer>
-		<layer id="6794" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="4313" name="Constant_17310" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6795" name="Constant_279436" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4314" name="Constant_17312" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6796" name="ScatterUpdate_279439" type="ScatterUpdate" version="opset3">
+		<layer id="4315" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Slice_2" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -103968,146 +65548,89 @@
 				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
 				<port id="4" precision="I64">
-					<dim>3</dim>
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="6797" name="Constant_279440" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="5" precision="FP32" names="3334,x2.105">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6798" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="4316" name="Constant_37501" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6799" name="ScatterUpdate_279441" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
 					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
 					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6800" name="Constant_279444" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438612" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
 			</output>
 		</layer>
-		<layer id="6801" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4317" name="__module.model.gpt_neox.layers.26.attention/aten::neg/Multiply" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="BOOL">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+				<port id="1" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>3</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="BOOL" names="4437">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="6802" name="Constant_279517" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="3335">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6803" name="Constant_279518" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4318" name="Constant_17205" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6804" name="Constant_279514" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4319" name="Constant_17249" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6805" name="Constant_279513" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4320" name="Constant_17251" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6806" name="ScatterUpdate_279519" type="ScatterUpdate" version="opset3">
+		<layer id="4321" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Slice_3" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -104115,183 +65638,128 @@
 				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
 				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6807" name="Constant_279522" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6808" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="BOOL">
 					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="BOOL" names="4438,4439">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="5" precision="FP32" names="3329">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6809" name="__module.model.gpt_neox.layers.27.attention/aten::where/Convert_1_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="6810" name="__module.model.gpt_neox.layers.27.attention/aten::where/Convert_1" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="6811" name="__module.model.gpt_neox.layers.27.attention/aten::full/Convert" type="Const" version="opset1">
-			<data element_type="f32" shape="" offset="205438636" size="4" />
-			<output>
-				<port id="0" precision="FP32" names="4426" />
-			</output>
-		</layer>
-		<layer id="6812" name="__module.model.gpt_neox.layers.27.attention/aten::where/Select" type="Select" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4322" name="__module.model.gpt_neox.layers.26.attention/aten::cat/Concat" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
-				<port id="1" precision="FP32" />
-				<port id="2" precision="FP32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="4440,causal_mask.109">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="3337">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6813" name="Add_290500" type="Add" version="opset1">
+		<layer id="4323" name="__module.model.gpt_neox.layers.26.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>1</dim>
 					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="3338">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6814" name="ShapeOf_290501" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="4324" name="__module.model.gpt_neox.layers.26.attention/aten::add/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="3339">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6815" name="Constant_254734" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4325" name="Constant_17160" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6816" name="__module.model.gpt_neox.layers.27.attention/aten::size/Constant_10" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4326" name="Constant_17162" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="6817" name="__module.model.gpt_neox.layers.27.attention/aten::size/Gather_10" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I32" />
-			</input>
+			</output>
+		</layer>
+		<layer id="4327" name="Constant_17164" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="3" precision="I64" names="4424">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6818" name="__module.model.gpt_neox.layers.27.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="4328" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Slice_4" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -104302,93 +65770,154 @@
 				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64" names="4441">
-					<dim>4</dim>
+				<port id="5" precision="FP32" names="3319">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6819" name="__module.model.gpt_neox.layers.27.attention/aten::expand/Abs" type="Abs" version="opset1">
+		<layer id="4329" name="__module.model.gpt_neox.layers.26.attention/aten::cat/Concat_1" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="3357">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6820" name="Maximum_290502" type="Maximum" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4330" name="Constant_53175" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
+			<output>
+				<port id="0" precision="FP32" />
+			</output>
+		</layer>
+		<layer id="4331" name="Constant_53168" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4332" name="Constant_53170" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4333" name="Constant_53172" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4334" name="Concat_53173" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
+				<port id="4" precision="I64">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6821" name="__module.model.gpt_neox.layers.27.attention/aten::add/Add_2" type="Broadcast" version="opset3">
+		<layer id="4335" name="Broadcast_53176" type="Broadcast" version="opset3">
 			<data mode="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
+				<port id="0" precision="FP32" />
 				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4443">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6822" name="__module.model.gpt_neox.layers.27.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
-			<data causal="false" />
+		<layer id="4336" name="ReadValue_50259" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.26.keypresent.26.key" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>-1</dim>
+					<dim>0</dim>
 					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
+			</input>
+			<output>
+				<port id="1" precision="FP32" names="past_key_values.26.key">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-				<port id="2" precision="FP32">
+			</output>
+		</layer>
+		<layer id="4337" name="Constant_47934" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64" />
+			</output>
+		</layer>
+		<layer id="4338" name="Gather_47935" type="Gather" version="opset8">
+			<data batch_dims="0" />
+			<input>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
+				<port id="1" precision="I32">
 					<dim>-1</dim>
 				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4444,4445,sdpa_result.55,tensor.109">
+				<port id="3" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -104396,28 +65925,53 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6823" name="Constant_16379" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="4339" name="Constant_17114" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
-				<port id="0" precision="I64" names="4446">
-					<dim>4</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4340" name="Constant_17116" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6824" name="__module.model.gpt_neox.layers.27.attention/aten::permute/Transpose_3" type="Transpose" version="opset1">
+		<layer id="4341" name="Constant_17118" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4342" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Slice_5" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4447">
+				<port id="5" precision="FP32" names="3312">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -104425,19 +65979,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6825" name="__module.model.gpt_neox.layers.27.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438650" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4343" name="Constant_17127" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="0" precision="I64" names="3313">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6826" name="__module.model.gpt_neox.layers.27.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
-			<data special_zero="true" />
+		<layer id="4344" name="__module.model.gpt_neox.layers.26.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -104446,986 +65996,960 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>3</dim>
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4452">
+				<port id="2" precision="FP32" names="3314,key.105">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6827" name="self.model.gpt_neox.layers.27.attention.dense.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 2560" offset="2332468434" size="6553600" />
+		<layer id="4345" name="Constant_17170" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6828" name="Convert_22514" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="4346" name="Constant_17172" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6829" name="self.model.gpt_neox.layers.27.attention.dense.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="2339022034" size="2560" />
+		<layer id="4347" name="Constant_17174" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6830" name="Convert_22517" type="Convert" version="opset1">
-			<data destination_type="f16" />
+		<layer id="4348" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Slice_6" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
+				<port id="5" precision="FP32" names="3320,k.53">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6831" name="self.model.gpt_neox.layers.27.attention.dense.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="4349" name="__module.model.gpt_neox.layers.26.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
 					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32" names="3340">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6832" name="self.model.gpt_neox.layers.27.attention.dense.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="2339024594" size="5120" />
+		<layer id="4350" name="Constant_17451" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6833" name="self.model.gpt_neox.layers.27.attention.dense.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+		<layer id="4351" name="Constant_17453" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="4352" name="Constant_17455" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6834" name="self.model.gpt_neox.layers.27.attention.dense.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="4353" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Slice_7" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="5" precision="FP32" names="3350,x2.107">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6835" name="__module.model.gpt_neox.layers.27.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
+		<layer id="4354" name="Constant_37502" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4355" name="__module.model.gpt_neox.layers.26.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>2560</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="3351">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6836" name="Constant_299348_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2339029714" size="5120" />
+		<layer id="4356" name="Constant_17348" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6837" name="Constant_299348" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+		<layer id="4357" name="Constant_17392" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="4358" name="Constant_17394" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6838" name="__module.model.gpt_neox.layers.27.attention.dense/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4359" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Slice_8" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="FP32">
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4455,4457,input.111">
+				<port id="5" precision="FP32" names="3345">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6839" name="__module.model.gpt_neox.layers.27/aten::add/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4360" name="__module.model.gpt_neox.layers.26.attention/aten::cat/Concat_2" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4475">
+				<port id="2" precision="FP32" names="3353">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6840" name="__module.model.gpt_neox.layers.27/aten::add/Add_1" type="Add" version="opset1">
+		<layer id="4361" name="__module.model.gpt_neox.layers.26.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4476,4478,hidden_states.57">
+				<port id="2" precision="FP32" names="3354">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6841" name="__module.model.gpt_neox.layers.28.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867670" size="4" />
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6842" name="__module.model.gpt_neox.layers.28.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
-			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
+		<layer id="4362" name="__module.model.gpt_neox.layers.26.attention/aten::add/Add_1" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="I32">
-					<dim>1</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="3355">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6843" name="Constant_299349_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2339034834" size="5120" />
+		<layer id="4363" name="Constant_17180" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6844" name="Constant_299349" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+		<layer id="4364" name="Constant_17182" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="4365" name="Constant_17184" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6845" name="__module.model.gpt_neox.layers.28.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4366" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Slice_9" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="1" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="6846" name="Constant_299350_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2339039954" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="2" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="6847" name="Constant_299350" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
+				<port id="3" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+				<port id="5" precision="FP32" names="3321">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6848" name="__module.model.gpt_neox.layers.28.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4367" name="__module.model.gpt_neox.layers.26.attention/aten::cat/Concat_3" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4613">
+				<port id="2" precision="FP32" names="3359">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6849" name="self.model.gpt_neox.layers.28.mlp.dense_h_to_4h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 2560" offset="2339045074" size="26214400" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6850" name="Convert_22564" type="Convert" version="opset1">
-			<data destination_type="f16" />
+		<layer id="4368" name="__module.model.gpt_neox.layers.26.attention/aten::cat/Concat_4" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32" names="present.26.key">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6851" name="self.model.gpt_neox.layers.28.mlp.dense_h_to_4h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 1" offset="2365259474" size="10240" />
+		<layer id="4369" name="Constant_53187" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
+				<port id="0" precision="FP32" />
 			</output>
 		</layer>
-		<layer id="6852" name="Convert_22567" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</input>
+		<layer id="4370" name="Constant_53180" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6853" name="self.model.gpt_neox.layers.28.mlp.dense_h_to_4h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</input>
+		<layer id="4371" name="Constant_53182" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="2" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6854" name="self.model.gpt_neox.layers.28.mlp.dense_h_to_4h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="10240, 1" offset="2365269714" size="20480" />
+		<layer id="4372" name="Constant_53184" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6855" name="self.model.gpt_neox.layers.28.mlp.dense_h_to_4h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4373" name="Concat_53185" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6856" name="self.model.gpt_neox.layers.28.mlp.dense_h_to_4h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="1" precision="I64">
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="6857" name="__module.model.gpt_neox.layers.28.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
+				<port id="2" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="3" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6858" name="Constant_299351_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 10240" offset="2365290194" size="20480" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
+				<port id="4" precision="I64">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6859" name="Constant_299351" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="4374" name="Broadcast_53188" type="Broadcast" version="opset3">
+			<data mode="numpy" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
+			<output>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6860" name="__module.model.gpt_neox.layers.28.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4375" name="ReadValue_50261" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.26.valuepresent.26.value" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>10240</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4618">
+				<port id="1" precision="FP32" names="past_key_values.26.value">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6861" name="__module.model.gpt_neox.layers.28.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
-			<data approximation_mode="ERF" />
+		<layer id="4376" name="Constant_47937" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64" />
+			</output>
+		</layer>
+		<layer id="4377" name="Gather_47938" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I32">
 					<dim>-1</dim>
-					<dim>10240</dim>
 				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="4619">
+				<port id="3" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6862" name="self.model.gpt_neox.layers.28.mlp.dense_4h_to_h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 10240" offset="2365310674" size="26214400" />
+		<layer id="4378" name="Constant_17132" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6863" name="Convert_22574" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
+		<layer id="4379" name="Constant_17134" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6864" name="self.model.gpt_neox.layers.28.mlp.dense_4h_to_h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="2391525074" size="2560" />
+		<layer id="4380" name="Constant_17136" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6865" name="Convert_22577" type="Convert" version="opset1">
-			<data destination_type="f16" />
+		<layer id="4381" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Slice_10" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>240</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="6866" name="self.model.gpt_neox.layers.28.mlp.dense_4h_to_h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="3" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="5" precision="FP32" names="3315">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6867" name="self.model.gpt_neox.layers.28.mlp.dense_4h_to_h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="2391527634" size="5120" />
+		<layer id="4382" name="Constant_17145" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
+				<port id="0" precision="I64" names="3316">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6868" name="self.model.gpt_neox.layers.28.mlp.dense_4h_to_h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4383" name="__module.model.gpt_neox.layers.26.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="2" precision="FP32" names="3317">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6869" name="self.model.gpt_neox.layers.28.mlp.dense_4h_to_h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="4384" name="__module.model.gpt_neox.layers.26.attention/aten::cat/Concat_5" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="2" precision="FP32" names="present.26.value">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6870" name="__module.model.gpt_neox.layers.28.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
+		<layer id="4385" name="Constant_17527" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4386" name="ShapeOf_40558" type="ShapeOf" version="opset3">
+			<data output_type="i64" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6871" name="Constant_299352_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2391532754" size="5120" />
+		<layer id="4387" name="Constant_40559" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116168" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6872" name="Constant_299352" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="4388" name="Constant_40560" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64" />
+			</output>
+		</layer>
+		<layer id="4389" name="Gather_40561" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
+					<dim>4</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="3" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6873" name="__module.model.gpt_neox.layers.28.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1">
+		<layer id="4390" name="Add_40564" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP32">
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4622,input.117">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
+				<port id="2" precision="I64" names="3364">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6874" name="Constant_299353_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2391537874" size="5120" />
+		<layer id="4391" name="Constant_17530" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6875" name="Constant_299353" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="4392" name="__module.model.gpt_neox.layers.26.attention/aten::slice/Slice_14" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
 					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="5" precision="FP32" names="3368">
+					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6876" name="__module.model.gpt_neox.layers.28.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4393" name="__module.model.gpt_neox.layers.26.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+			<data causal="false" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="3" precision="FP32">
+					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="4" precision="FP32" names="3369,attn_output.157">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6877" name="Constant_299354_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2391542994" size="5120" />
+		<layer id="4394" name="__module.model.gpt_neox.layers.26.attention/aten::transpose/Constant" type="Const" version="opset1">
+			<data element_type="i32" shape="4" offset="201116200" size="16" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I32">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6878" name="Constant_299354" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="4395" name="__module.model.gpt_neox.layers.26.attention/aten::transpose/Transpose" type="Transpose" version="opset1">
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I32">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32" names="3370">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6879" name="__module.model.gpt_neox.layers.28.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4396" name="Constant_41010" type="Const" version="opset1">
+			<data element_type="i64" shape="3" offset="201116216" size="24" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>3</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4397" name="__module.model.gpt_neox.layers.26.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
+			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+				<port id="1" precision="I64">
+					<dim>3</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4488">
+				<port id="2" precision="FP32" names="3373">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6880" name="self.model.gpt_neox.layers.28.attention.query_key_value.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 2560" offset="2391548114" size="19660800" />
+		<layer id="4398" name="self.model.gpt_neox.layers.26.attention.dense.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 2560" offset="2250432080" size="6553600" />
 			<output>
 				<port id="0" precision="U8">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6881" name="Convert_22544" type="Convert" version="opset1">
+		<layer id="4399" name="Convert_75462" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP16">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6882" name="self.model.gpt_neox.layers.28.attention.query_key_value.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 1" offset="2411208914" size="7680" />
+		<layer id="4400" name="self.model.gpt_neox.layers.26.attention.dense.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="2256985680" size="2560" />
 			<output>
 				<port id="0" precision="U8">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6883" name="Convert_22547" type="Convert" version="opset1">
+		<layer id="4401" name="Convert_75465" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP16">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6884" name="self.model.gpt_neox.layers.28.attention.query_key_value.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="4402" name="self.model.gpt_neox.layers.26.attention.dense.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP16">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
 				<port id="2" precision="FP16">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6885" name="self.model.gpt_neox.layers.28.attention.query_key_value.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="7680, 1" offset="2411216594" size="15360" />
+		<layer id="4403" name="self.model.gpt_neox.layers.26.attention.dense.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="2256988240" size="5120" />
 			<output>
 				<port id="0" precision="FP16">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6886" name="self.model.gpt_neox.layers.28.attention.query_key_value.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="4404" name="self.model.gpt_neox.layers.26.attention.dense.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP16">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
 				<port id="2" precision="FP16">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6887" name="self.model.gpt_neox.layers.28.attention.query_key_value.weight" type="Convert" version="opset1">
+		<layer id="4405" name="self.model.gpt_neox.layers.26.attention.dense.weight/fq_weights_1/convert" type="Convert" version="opset1">
 			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
 			<input>
 				<port id="0" precision="FP16">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP32">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6888" name="__module.model.gpt_neox.layers.28.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="4406" name="__module.model.gpt_neox.layers.26.attention.dense/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -105434,7 +66958,7 @@
 					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>7680</dim>
+					<dim>2560</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
@@ -105442,826 +66966,838 @@
 				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6889" name="Constant_299355_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 7680" offset="2411231954" size="15360" />
+		<layer id="4407" name="Constant_37503" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2256993360" size="10240" />
 			<output>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6890" name="Constant_299355" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="4408" name="__module.model.gpt_neox.layers.26.attention.dense/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
+				<port id="2" precision="FP32" names="3376,3378,input.105">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6891" name="__module.model.gpt_neox.layers.28.attention.query_key_value/aten::linear/Add" type="Add" version="opset1">
+		<layer id="4409" name="__module.model.gpt_neox.layers.26/aten::add/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4495,qkv.113">
+				<port id="2" precision="FP32" names="3396">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6892" name="__module.model.gpt_neox.layers.28.attention/prim::ListConstruct/Concat" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080154" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6893" name="__module.model.gpt_neox.layers.28.attention/aten::view/Reshape" type="Reshape" version="opset1">
-			<data special_zero="true" />
+		<layer id="4410" name="__module.model.gpt_neox.layers.26/aten::add/Add_1" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4499,qkv.115">
+				<port id="2" precision="FP32" names="3397,3399,hidden_states.111">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6894" name="Constant_279529" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6895" name="Constant_279532" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6896" name="Constant_279535" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4411" name="__module.model.gpt_neox.layers.27.input_layernorm/aten::layer_norm/Multiply" type="Const" version="opset1">
+			<data element_type="i32" shape="1" offset="128867668" size="4" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="I32">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6897" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4412" name="__module.model.gpt_neox.layers.27.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
+			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="I32">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4500">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6898" name="Constant_16491" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="4413" name="Constant_37504" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2257003600" size="10240" />
 			<output>
-				<port id="0" precision="I64" names="4501">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6899" name="__module.model.gpt_neox.layers.28.attention/aten::permute/Transpose" type="Transpose" version="opset1">
+		<layer id="4414" name="__module.model.gpt_neox.layers.27.post_attention_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4502,query.169">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6900" name="Constant_279541" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4415" name="Constant_37505" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2257013840" size="10240" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6901" name="Constant_279544" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4416" name="__module.model.gpt_neox.layers.27.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="3497">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6902" name="Constant_279547" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4417" name="self.model.gpt_neox.layers.27.mlp.dense_h_to_4h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 2560" offset="2257024080" size="26214400" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6903" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4418" name="Convert_74724" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4509,q.57">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6904" name="self.model.gpt_neox.layers.28.attention.rotary_emb.cos_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201080346" size="81920" />
+		<layer id="4419" name="self.model.gpt_neox.layers.27.mlp.dense_h_to_4h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 1" offset="2283238480" size="10240" />
 			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.28.attention.rotary_emb.cos_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6905" name="self.model.gpt_neox.layers.28.attention.rotary_emb.cos_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="4420" name="Convert_74727" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6906" name="__module.model.gpt_neox.layers.28.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
+		<layer id="4421" name="self.model.gpt_neox.layers.27.mlp.dense_h_to_4h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6907" name="Constant_279553" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4422" name="self.model.gpt_neox.layers.27.mlp.dense_h_to_4h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="10240, 1" offset="2283248720" size="20480" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6908" name="Constant_279556" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4423" name="self.model.gpt_neox.layers.27.mlp.dense_h_to_4h.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6909" name="Constant_279559" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
+		<layer id="4424" name="__module.model.gpt_neox.layers.27.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<rt_info>
-				<attribute name="precise" version="0" />
+				<attribute name="decompression" version="0" />
 			</rt_info>
+			<input>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6910" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4425" name="__module.model.gpt_neox.layers.27.mlp.dense_h_to_4h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4503">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6911" name="Constant_16505" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="4426" name="self.model.gpt_neox.layers.27.mlp.dense_h_to_4h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="10240" offset="2283269200" size="20480" />
 			<output>
-				<port id="0" precision="I64" names="4504">
-					<dim>4</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.27.mlp.dense_h_to_4h.bias">
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4427" name="__module.model.gpt_neox.layers.27.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<rt_info>
+				<attribute name="decompression" version="0" />
+			</rt_info>
+			<input>
+				<port id="0" precision="BF16">
+					<dim>10240</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6912" name="__module.model.gpt_neox.layers.28.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
+		<layer id="4428" name="__module.model.gpt_neox.layers.27.mlp.dense_h_to_4h/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4505,key.169">
+				<port id="2" precision="FP32" names="3502">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6913" name="__module.model.gpt_neox.layers.28.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="4429" name="__module.model.gpt_neox.layers.27.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
+			<data approximation_mode="ERF" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32" names="3503">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6914" name="__module.model.gpt_neox.layers.28.attention/aten::size/Constant_2" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4430" name="self.model.gpt_neox.layers.27.mlp.dense_4h_to_h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 10240" offset="2283289680" size="26214400" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6915" name="__module.model.gpt_neox.layers.28.attention/aten::size/Gather_2" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="4431" name="Convert_74734" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="4513" />
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6916" name="Constant_338157_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
+		<layer id="4432" name="self.model.gpt_neox.layers.27.mlp.dense_4h_to_h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="2309504080" size="2560" />
 			<output>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6917" name="Constant_338157" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="4433" name="Convert_74737" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="6918" name="Constant_338150" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
-			<output>
-				<port id="0" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6919" name="Constant_338152" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64">
+		<layer id="4434" name="self.model.gpt_neox.layers.27.mlp.dense_4h_to_h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6920" name="Constant_338154" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
+		<layer id="4435" name="self.model.gpt_neox.layers.27.mlp.dense_4h_to_h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="2309506640" size="5120" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6921" name="Concat_338155" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="4436" name="self.model.gpt_neox.layers.27.mlp.dense_4h_to_h.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="3" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6922" name="Broadcast_338158" type="Broadcast" version="opset3">
-			<data mode="numpy" />
+		<layer id="4437" name="__module.model.gpt_neox.layers.27.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<rt_info>
+				<attribute name="decompression" version="0" />
+			</rt_info>
 			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6923" name="ReadValue_330509" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.28.keypresent.28.key" variable_type="f32" variable_shape="?,32,?,80" />
+		<layer id="4438" name="__module.model.gpt_neox.layers.27.mlp.dense_4h_to_h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="past_key_values.28.key">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6924" name="Constant_323553" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
+		<layer id="4439" name="self.model.gpt_neox.layers.27.mlp.dense_4h_to_h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="2560" offset="2309511760" size="5120" />
 			<output>
-				<port id="0" precision="I64" />
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.27.mlp.dense_4h_to_h.bias">
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6925" name="Gather_323554" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="4440" name="__module.model.gpt_neox.layers.27.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<rt_info>
+				<attribute name="decompression" version="0" />
+			</rt_info>
+			<input>
+				<port id="0" precision="BF16">
+					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4441" name="__module.model.gpt_neox.layers.27.mlp.dense_4h_to_h/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
 				</port>
-				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="3" precision="FP32">
+				<port id="2" precision="FP32" names="3506,input.111">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6926" name="__module.model.gpt_neox.layers.28.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="4442" name="Constant_37506" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2309516880" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4443" name="__module.model.gpt_neox.layers.27.input_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6927" name="__module.model.gpt_neox.layers.28.attention/aten::size/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4444" name="Constant_37507" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2309527120" size="10240" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="6928" name="__module.model.gpt_neox.layers.28.attention/aten::size/Gather_3" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="4515" />
 			</output>
 		</layer>
-		<layer id="6929" name="__module.model.gpt_neox.layers.28.attention/aten::add_/Add" type="Add" version="opset1">
+		<layer id="4445" name="__module.model.gpt_neox.layers.27.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64" names="4514,4518,4519,seq_len.113" />
+				<port id="2" precision="FP32" names="3409,hidden_states.109">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6930" name="__module.model.gpt_neox.layers.28.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4446" name="self.model.gpt_neox.layers.27.attention.query_key_value.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 2560" offset="2309537360" size="19660800" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6931" name="__module.model.gpt_neox.layers.28.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1">
+		<layer id="4447" name="Convert_75121" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6932" name="__module.model.gpt_neox.layers.28.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4448" name="self.model.gpt_neox.layers.27.attention.query_key_value.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 1" offset="2329198160" size="7680" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="U8">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6933" name="__module.model.gpt_neox.layers.28.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4449" name="Convert_75124" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
+				<port id="0" precision="U8">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
+			</input>
+			<output>
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I64">
+			</output>
+		</layer>
+		<layer id="4450" name="self.model.gpt_neox.layers.27.attention.query_key_value.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4522,4525,cos.113">
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="2" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6934" name="Constant_38365" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
+		<layer id="4451" name="self.model.gpt_neox.layers.27.attention.query_key_value.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="7680, 1" offset="2329205840" size="15360" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6935" name="__module.model.gpt_neox.layers.28.attention/aten::index/Gather" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="4452" name="self.model.gpt_neox.layers.27.attention.query_key_value.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
+					<dim>1</dim>
 				</port>
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="4528">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="2" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6936" name="__module.model.gpt_neox.layers.28.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1">
+		<layer id="4453" name="self.model.gpt_neox.layers.27.attention.query_key_value.weight/fq_weights_1/convert" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4529,cos.115">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="1" precision="FP32">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6937" name="__module.model.gpt_neox.layers.28.attention/aten::mul/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4454" name="__module.model.gpt_neox.layers.27.attention.query_key_value/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4533">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6938" name="Constant_279564" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6939" name="Constant_279563" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4455" name="Constant_37508" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 7680" offset="2329221200" size="30720" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="FP32">
+					<dim>1</dim>
 					<dim>1</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6940" name="__module.model.gpt_neox.layers.28.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="4456" name="__module.model.gpt_neox.layers.27.attention.query_key_value/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>7680</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>7680</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="3416,qkv.109">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6941" name="__module.model.gpt_neox.layers.28.attention/aten::size/Constant_5" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4457" name="__module.model.gpt_neox.layers.27.attention/prim::ListConstruct" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201115992" size="32" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="6942" name="__module.model.gpt_neox.layers.28.attention/aten::size/Gather_5" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
 				<port id="0" precision="I64">
 					<dim>4</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="4534,4539" />
-			</output>
-		</layer>
-		<layer id="6943" name="__module.model.gpt_neox.layers.28.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" />
 			</output>
 		</layer>
-		<layer id="6944" name="__module.model.gpt_neox.layers.28.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1">
+		<layer id="4458" name="__module.model.gpt_neox.layers.27.attention/aten::view/Reshape" type="Reshape" version="opset1">
+			<data special_zero="true" />
 			<input>
-				<port id="0" precision="I64" />
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>7680</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>4</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64" names="4536,4537,4541,4542" />
+				<port id="2" precision="FP32" names="3420,qkv.111">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>240</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6945" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Constant_1" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4459" name="Constant_17726" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6946" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="4460" name="Constant_17728" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6947" name="Constant_279562" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4461" name="Constant_17730" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6948" name="ScatterUpdate_279565" type="ScatterUpdate" version="opset3">
+		<layer id="4462" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Slice" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -106269,148 +67805,155 @@
 				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
 				<port id="4" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="6949" name="Constant_279568" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="5" precision="FP32" names="3421">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6950" name="Constant_279571" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4463" name="Constant_17739" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="I64" names="3422">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6951" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4464" name="__module.model.gpt_neox.layers.27.attention/aten::permute/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4543,x2.113">
+				<port id="2" precision="FP32" names="3423,query.55">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6952" name="Constant_299356_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
+		<layer id="4465" name="Constant_17780" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4466" name="Constant_17782" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4467" name="Constant_17784" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6953" name="Constant_299356" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="4468" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Slice_1" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="5" precision="FP32" names="3430,q.55">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6954" name="__module.model.gpt_neox.layers.28.attention/aten::neg/Multiply" type="Multiply" version="opset1">
+		<layer id="4469" name="__module.model.gpt_neox.layers.27.attention/aten::mul/Multiply" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
+					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4544">
+				<port id="2" precision="FP32" names="3436">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6955" name="Constant_279616" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4470" name="Constant_17938" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6956" name="Constant_279621" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4471" name="Constant_17940" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4472" name="Constant_17942" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6957" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4473" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Slice_2" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -106419,17 +67962,20 @@
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4538">
+				<port id="5" precision="FP32" names="3446,x2.109">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -106437,8 +67983,19 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6958" name="__module.model.gpt_neox.layers.28.attention/aten::cat/Concat" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="4474" name="Constant_37509" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4475" name="__module.model.gpt_neox.layers.27.attention/aten::neg/Multiply" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -106447,75 +68004,51 @@
 					<dim>10</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4546">
+				<port id="2" precision="FP32" names="3447">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6959" name="self.model.gpt_neox.layers.28.attention.rotary_emb.sin_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201162356" size="81920" />
-			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.28.attention.rotary_emb.sin_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6960" name="self.model.gpt_neox.layers.28.attention.rotary_emb.sin_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</input>
+		<layer id="4476" name="Constant_17835" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6961" name="__module.model.gpt_neox.layers.28.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4477" name="Constant_17879" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6962" name="__module.model.gpt_neox.layers.28.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4478" name="Constant_17881" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6963" name="__module.model.gpt_neox.layers.28.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4479" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Slice_3" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>2048</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
@@ -106527,60 +68060,45 @@
 				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4523,4526,sin.113">
+				<port id="5" precision="FP32" names="3441">
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6964" name="Constant_38368" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="6965" name="__module.model.gpt_neox.layers.28.attention/aten::index/Gather_1" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="4480" name="__module.model.gpt_neox.layers.27.attention/aten::cat/Concat" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="4531">
-					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>10</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="6966" name="__module.model.gpt_neox.layers.28.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
+				<port id="1" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>10</dim>
 				</port>
-				<port id="1" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4532,sin.115">
+				<port id="2" precision="FP32" names="3449">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6967" name="__module.model.gpt_neox.layers.28.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
+		<layer id="4481" name="__module.model.gpt_neox.layers.27.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -106597,7 +68115,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4547">
+				<port id="2" precision="FP32" names="3450">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -106605,7 +68123,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6968" name="__module.model.gpt_neox.layers.28.attention/aten::add/Add" type="Add" version="opset1">
+		<layer id="4482" name="__module.model.gpt_neox.layers.27.attention/aten::add/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -106622,7 +68140,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4548">
+				<port id="2" precision="FP32" names="3451">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -106630,41 +68148,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6969" name="Constant_279667" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4483" name="Constant_17790" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6970" name="Constant_279670" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4484" name="Constant_17792" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6971" name="Constant_279673" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4485" name="Constant_17794" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6972" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4486" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Slice_4" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -106673,17 +68181,20 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4510">
+				<port id="5" precision="FP32" names="3431">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -106691,7 +68202,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6973" name="__module.model.gpt_neox.layers.28.attention/aten::cat/Concat_1" type="Concat" version="opset1">
+		<layer id="4487" name="__module.model.gpt_neox.layers.27.attention/aten::cat/Concat_1" type="Concat" version="opset1">
 			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
@@ -106708,7 +68219,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4566,4576,query.171,query.173">
+				<port id="2" precision="FP32" names="3469">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -106716,205 +68227,237 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6974" name="Constant_279679" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4488" name="Constant_53199" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
+			<output>
+				<port id="0" precision="FP32" />
+			</output>
+		</layer>
+		<layer id="4489" name="Constant_53192" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6975" name="Constant_279682" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4490" name="Constant_53194" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6976" name="Constant_279685" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4491" name="Constant_53196" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6977" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4492" name="Concat_53197" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="4" precision="I64">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4493" name="Broadcast_53200" type="Broadcast" version="opset3">
+			<data mode="numpy" />
+			<input>
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4511,k.57">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6978" name="__module.model.gpt_neox.layers.28.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4494" name="ReadValue_50263" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.27.keypresent.27.key" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
 				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP32" names="past_key_values.27.key">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
+			</output>
+		</layer>
+		<layer id="4495" name="Constant_47940" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64" />
+			</output>
+		</layer>
+		<layer id="4496" name="Gather_47941" type="Gather" version="opset8">
+			<data batch_dims="0" />
+			<input>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
+				<port id="1" precision="I32">
+					<dim>-1</dim>
+				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4549">
+				<port id="3" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6979" name="Constant_279689" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4497" name="Constant_17744" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6980" name="Constant_279688" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4498" name="Constant_17746" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6981" name="__module.model.gpt_neox.layers.28.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="4499" name="Constant_17748" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4500" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Slice_5" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
-			</input>
-			<output>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="6982" name="__module.model.gpt_neox.layers.28.attention/aten::size/Constant_7" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="6983" name="__module.model.gpt_neox.layers.28.attention/aten::size/Gather_7" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="4550,4555" />
+				<port id="5" precision="FP32" names="3424">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6984" name="__module.model.gpt_neox.layers.28.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
+		<layer id="4501" name="Constant_17757" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="2" precision="I64" />
+				<port id="0" precision="I64" names="3425">
+					<dim>4</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6985" name="__module.model.gpt_neox.layers.28.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1">
+		<layer id="4502" name="__module.model.gpt_neox.layers.27.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
 			<input>
-				<port id="0" precision="I64" />
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>4</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64" names="4552,4553,4557,4558" />
+				<port id="2" precision="FP32" names="3426,key.109">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6986" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4503" name="Constant_17800" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="6987" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="4504" name="Constant_17802" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6988" name="Constant_279687" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4505" name="Constant_17804" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6989" name="ScatterUpdate_279690" type="ScatterUpdate" version="opset3">
+		<layer id="4506" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Slice_6" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -106922,40 +68465,72 @@
 				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="5" precision="FP32" names="3432,k.55">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6990" name="Constant_279693" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4507" name="__module.model.gpt_neox.layers.27.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="3452">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4508" name="Constant_18081" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6991" name="Constant_279696" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4509" name="Constant_18083" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4510" name="Constant_18085" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6992" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4511" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Slice_7" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -106964,17 +68539,20 @@
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4559,x2.115">
+				<port id="5" precision="FP32" names="3462,x2.111">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -106982,32 +68560,10 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6993" name="Constant_299357_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="6994" name="Constant_299357" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
+		<layer id="4512" name="Constant_37510" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>1</dim>
@@ -107015,7 +68571,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6995" name="__module.model.gpt_neox.layers.28.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
+		<layer id="4513" name="__module.model.gpt_neox.layers.27.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -107032,7 +68588,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4560">
+				<port id="2" precision="FP32" names="3463">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -107040,30 +68596,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6996" name="Constant_279741" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4514" name="Constant_17978" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6997" name="Constant_279746" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4515" name="Constant_18022" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="6998" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4516" name="Constant_18024" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4517" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Slice_8" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -107072,17 +68629,20 @@
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4554">
+				<port id="5" precision="FP32" names="3457">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -107090,7 +68650,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="6999" name="__module.model.gpt_neox.layers.28.attention/aten::cat/Concat_2" type="Concat" version="opset1">
+		<layer id="4518" name="__module.model.gpt_neox.layers.27.attention/aten::cat/Concat_2" type="Concat" version="opset1">
 			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
@@ -107107,7 +68667,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4562">
+				<port id="2" precision="FP32" names="3465">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -107115,7 +68675,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7000" name="__module.model.gpt_neox.layers.28.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
+		<layer id="4519" name="__module.model.gpt_neox.layers.27.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -107132,7 +68692,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4563">
+				<port id="2" precision="FP32" names="3466">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -107140,7 +68700,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7001" name="__module.model.gpt_neox.layers.28.attention/aten::add/Add_1" type="Add" version="opset1">
+		<layer id="4520" name="__module.model.gpt_neox.layers.27.attention/aten::add/Add_1" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -107157,7 +68717,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4564">
+				<port id="2" precision="FP32" names="3467">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -107165,41 +68725,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7002" name="Constant_279791" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4521" name="Constant_17810" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7003" name="Constant_279794" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4522" name="Constant_17812" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7004" name="Constant_279797" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4523" name="Constant_17814" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7005" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4524" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Slice_9" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -107208,17 +68758,20 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4512">
+				<port id="5" precision="FP32" names="3433">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -107226,7 +68779,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7006" name="__module.model.gpt_neox.layers.28.attention/aten::cat/Concat_3" type="Concat" version="opset1">
+		<layer id="4525" name="__module.model.gpt_neox.layers.27.attention/aten::cat/Concat_3" type="Concat" version="opset1">
 			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
@@ -107243,7 +68796,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4568">
+				<port id="2" precision="FP32" names="3471">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -107251,7 +68804,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7007" name="__module.model.gpt_neox.layers.28.attention/aten::to/Convert_1" type="Concat" version="opset1">
+		<layer id="4526" name="__module.model.gpt_neox.layers.27.attention/aten::cat/Concat_4" type="Concat" version="opset1">
 			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
@@ -107268,7 +68821,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="present.28.key">
+				<port id="2" precision="FP32" names="present.27.key">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -107276,49 +68829,37 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7008" name="Constant_338170_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="7009" name="Constant_338170" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
+		<layer id="4527" name="Constant_53211" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="1" precision="FP32" />
+				<port id="0" precision="FP32" />
 			</output>
 		</layer>
-		<layer id="7010" name="Constant_338163" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
+		<layer id="4528" name="Constant_53204" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7011" name="Constant_338165" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
+		<layer id="4529" name="Constant_53206" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7012" name="Constant_338167" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
+		<layer id="4530" name="Constant_53208" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7013" name="Concat_338168" type="Concat" version="opset1">
+		<layer id="4531" name="Concat_53209" type="Concat" version="opset1">
 			<data axis="0" />
 			<input>
 				<port id="0" precision="I64">
@@ -107340,7 +68881,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7014" name="Broadcast_338171" type="Broadcast" version="opset3">
+		<layer id="4532" name="Broadcast_53212" type="Broadcast" version="opset3">
 			<data mode="numpy" />
 			<input>
 				<port id="0" precision="FP32" />
@@ -107357,8 +68898,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7015" name="ReadValue_330511" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.28.valuepresent.28.value" variable_type="f32" variable_shape="?,32,?,80" />
+		<layer id="4533" name="ReadValue_50265" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.27.valuepresent.27.value" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -107368,7 +68909,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="past_key_values.28.value">
+				<port id="1" precision="FP32" names="past_key_values.27.value">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -107376,13 +68917,13 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7016" name="Constant_323556" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
+		<layer id="4534" name="Constant_47943" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="7017" name="Gather_323557" type="Gather" version="opset8">
+		<layer id="4535" name="Gather_47944" type="Gather" version="opset8">
 			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="FP32">
@@ -107405,41 +68946,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7018" name="Constant_279803" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4536" name="Constant_17762" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7019" name="Constant_279806" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4537" name="Constant_17764" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7020" name="Constant_279809" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4538" name="Constant_17766" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7021" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4539" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Slice_10" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -107448,17 +68979,20 @@
 					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4506">
+				<port id="5" precision="FP32" names="3427">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -107466,15 +69000,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7022" name="Constant_16519" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="4540" name="Constant_17775" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64" names="4507">
+				<port id="0" precision="I64" names="3428">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7023" name="__module.model.gpt_neox.layers.28.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
+		<layer id="4541" name="__module.model.gpt_neox.layers.27.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -107487,7 +69021,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4508">
+				<port id="2" precision="FP32" names="3429">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -107495,7 +69029,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7024" name="__module.model.gpt_neox.layers.28.attention/aten::cat/Concat_5" type="Concat" version="opset1">
+		<layer id="4542" name="__module.model.gpt_neox.layers.27.attention/aten::cat/Concat_5" type="Concat" version="opset1">
 			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
@@ -107512,7 +69046,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="present.28.value">
+				<port id="2" precision="FP32" names="present.27.value">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -107520,78 +69054,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7025" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Slice_11" type="Const" version="opset1">
-			<data element_type="boolean" shape="1, 1, 2048, 2048" offset="201244276" size="4194304" />
-			<output>
-				<port id="0" precision="BOOL" names="4585">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7026" name="Constant_279813" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7027" name="Constant_279812" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162298" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4543" name="Constant_18157" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7028" name="__module.model.gpt_neox.layers.28.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7029" name="__module.model.gpt_neox.layers.28.attention/aten::size/Constant_8" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="7030" name="__module.model.gpt_neox.layers.28.attention/aten::size/Gather_8" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="4580" />
-			</output>
-		</layer>
-		<layer id="7031" name="__module.model.gpt_neox.layers.28.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3">
+		<layer id="4544" name="ShapeOf_40629" type="ShapeOf" version="opset3">
 			<data output_type="i64" />
 			<input>
 				<port id="0" precision="FP32">
@@ -107607,348 +69078,62 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7032" name="__module.model.gpt_neox.layers.28.attention/aten::size/Constant_9" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="7033" name="__module.model.gpt_neox.layers.28.attention/aten::size/Gather_9" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="4578" />
-			</output>
-		</layer>
-		<layer id="7034" name="__module.model.gpt_neox.layers.28.attention/aten::sub/Subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" names="4582,4583" />
-			</output>
-		</layer>
-		<layer id="7035" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Constant_4" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="7036" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7037" name="Constant_279811" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4545" name="Constant_40630" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116168" size="8" />
 			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7038" name="ScatterUpdate_279814" type="ScatterUpdate" version="opset3">
-			<input>
 				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
-				</port>
 			</output>
 		</layer>
-		<layer id="7039" name="Constant_279815" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4546" name="Constant_40631" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7040" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1">
-			<input>
 				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
 			</output>
 		</layer>
-		<layer id="7041" name="ScatterUpdate_279816" type="ScatterUpdate" version="opset3">
+		<layer id="4547" name="Gather_40632" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="I64">
-					<dim>3</dim>
+					<dim>4</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7042" name="Constant_279819" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438612" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7043" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>3</dim>
-				</port>
 				<port id="3" precision="I64">
-					<dim>3</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="BOOL" names="4586">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7044" name="Constant_279892" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7045" name="Constant_279893" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7046" name="Constant_279889" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7047" name="Constant_279888" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7048" name="ScatterUpdate_279894" type="ScatterUpdate" version="opset3">
+		<layer id="4548" name="Add_40635" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7049" name="Constant_279897" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7050" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="BOOL" names="4587,4588">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7051" name="__module.model.gpt_neox.layers.28.attention/aten::where/Convert_1_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="7052" name="__module.model.gpt_neox.layers.28.attention/aten::where/Convert_1" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="7053" name="__module.model.gpt_neox.layers.28.attention/aten::full/Convert" type="Const" version="opset1">
-			<data element_type="f32" shape="" offset="205438636" size="4" />
-			<output>
-				<port id="0" precision="FP32" names="4575" />
-			</output>
-		</layer>
-		<layer id="7054" name="__module.model.gpt_neox.layers.28.attention/aten::where/Select" type="Select" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
 				</port>
-				<port id="1" precision="FP32" />
-				<port id="2" precision="FP32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="4589,causal_mask.113">
+				<port id="2" precision="I64" names="3476">
 					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7055" name="Add_290506" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-				</port>
-			</input>
+		<layer id="4549" name="Constant_18160" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7056" name="ShapeOf_290507" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="4550" name="__module.model.gpt_neox.layers.27.attention/aten::slice/Slice_14" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -107956,56 +69141,6 @@
 					<dim>-1</dim>
 					<dim>-1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7057" name="Constant_254749" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7058" name="__module.model.gpt_neox.layers.28.attention/aten::size/Constant_10" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="7059" name="__module.model.gpt_neox.layers.28.attention/aten::size/Gather_10" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="4573">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7060" name="__module.model.gpt_neox.layers.28.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1">
-			<data axis="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
@@ -108015,64 +69150,20 @@
 				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64" names="4590">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7061" name="__module.model.gpt_neox.layers.28.attention/aten::expand/Abs" type="Abs" version="opset1">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7062" name="Maximum_290508" type="Maximum" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7063" name="__module.model.gpt_neox.layers.28.attention/aten::add/Add_2" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4592">
-					<dim>-1</dim>
+				<port id="5" precision="FP32" names="3480">
 					<dim>-1</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
 					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7064" name="__module.model.gpt_neox.layers.28.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+		<layer id="4551" name="__module.model.gpt_neox.layers.27.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
 			<data causal="false" />
 			<input>
 				<port id="0" precision="FP32">
@@ -108095,13 +69186,13 @@
 				</port>
 				<port id="3" precision="FP32">
 					<dim>-1</dim>
-					<dim>-1</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
 					<dim>-1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4593,4594,sdpa_result.57,tensor.113">
+				<port id="4" precision="FP32" names="3481,attn_output.163">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -108109,15 +69200,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7065" name="Constant_16960" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="4552" name="__module.model.gpt_neox.layers.27.attention/aten::transpose/Constant" type="Const" version="opset1">
+			<data element_type="i32" shape="4" offset="201116200" size="16" />
 			<output>
-				<port id="0" precision="I64" names="4595">
+				<port id="0" precision="I32">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7066" name="__module.model.gpt_neox.layers.28.attention/aten::permute/Transpose_3" type="Transpose" version="opset1">
+		<layer id="4553" name="__module.model.gpt_neox.layers.27.attention/aten::transpose/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -108125,12 +69216,12 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="I32">
 					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4596">
+				<port id="2" precision="FP32" names="3482">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -108138,18 +69229,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7067" name="__module.model.gpt_neox.layers.28.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438650" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4554" name="Constant_41011" type="Const" version="opset1">
+			<data element_type="i64" shape="3" offset="201116216" size="24" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>3</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7068" name="__module.model.gpt_neox.layers.28.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
+		<layer id="4555" name="__module.model.gpt_neox.layers.27.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
 			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -108163,15 +69251,15 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4601">
+				<port id="2" precision="FP32" names="3485">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7069" name="self.model.gpt_neox.layers.28.attention.dense.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 2560" offset="2411247314" size="6553600" />
+		<layer id="4556" name="self.model.gpt_neox.layers.27.attention.dense.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 2560" offset="2329251920" size="6553600" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>2560</dim>
@@ -108179,7 +69267,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7070" name="Convert_22554" type="Convert" version="opset1">
+		<layer id="4557" name="Convert_75473" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -108194,8 +69282,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7071" name="self.model.gpt_neox.layers.28.attention.dense.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="2417800914" size="2560" />
+		<layer id="4558" name="self.model.gpt_neox.layers.27.attention.dense.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="2335805520" size="2560" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>2560</dim>
@@ -108203,7 +69291,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7072" name="Convert_22557" type="Convert" version="opset1">
+		<layer id="4559" name="Convert_75476" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -108218,7 +69306,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7073" name="self.model.gpt_neox.layers.28.attention.dense.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="4560" name="self.model.gpt_neox.layers.27.attention.dense.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -108237,8 +69325,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7074" name="self.model.gpt_neox.layers.28.attention.dense.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="2417803474" size="5120" />
+		<layer id="4561" name="self.model.gpt_neox.layers.27.attention.dense.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="2335808080" size="5120" />
 			<output>
 				<port id="0" precision="FP16">
 					<dim>2560</dim>
@@ -108246,7 +69334,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7075" name="self.model.gpt_neox.layers.28.attention.dense.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="4562" name="self.model.gpt_neox.layers.27.attention.dense.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -108265,11 +69353,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7076" name="self.model.gpt_neox.layers.28.attention.dense.weight" type="Convert" version="opset1">
+		<layer id="4563" name="self.model.gpt_neox.layers.27.attention.dense.weight/fq_weights_1/convert" type="Convert" version="opset1">
 			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
 			<input>
 				<port id="0" precision="FP16">
 					<dim>2560</dim>
@@ -108283,7 +69368,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7077" name="__module.model.gpt_neox.layers.28.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="4564" name="__module.model.gpt_neox.layers.27.attention.dense/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -108304,37 +69389,17 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7078" name="Constant_299358_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2417808594" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7079" name="Constant_299358" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="4565" name="Constant_37511" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2335813200" size="10240" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7080" name="__module.model.gpt_neox.layers.28.attention.dense/aten::linear/Add" type="Add" version="opset1">
+		<layer id="4566" name="__module.model.gpt_neox.layers.27.attention.dense/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -108349,14 +69414,14 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4604,4606,input.115">
+				<port id="2" precision="FP32" names="3488,3490,input.109">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7081" name="__module.model.gpt_neox.layers.28/aten::add/Add" type="Add" version="opset1">
+		<layer id="4567" name="__module.model.gpt_neox.layers.27/aten::add/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -108371,14 +69436,14 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4624">
+				<port id="2" precision="FP32" names="3508">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7082" name="__module.model.gpt_neox.layers.28/aten::add/Add_1" type="Add" version="opset1">
+		<layer id="4568" name="__module.model.gpt_neox.layers.27/aten::add/Add_1" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -108393,22 +69458,22 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4625,4627,hidden_states.59">
+				<port id="2" precision="FP32" names="3509,3511,hidden_states.115">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7083" name="__module.model.gpt_neox.layers.29.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867670" size="4" />
+		<layer id="4569" name="__module.model.gpt_neox.layers.28.input_layernorm/aten::layer_norm/Multiply" type="Const" version="opset1">
+			<data element_type="i32" shape="1" offset="128867668" size="4" />
 			<output>
 				<port id="0" precision="I32">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7084" name="__module.model.gpt_neox.layers.29.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
+		<layer id="4570" name="__module.model.gpt_neox.layers.28.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
 			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
 			<input>
 				<port id="0" precision="FP32">
@@ -108428,37 +69493,17 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7085" name="Constant_299359_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2417813714" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7086" name="Constant_299359" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="4571" name="Constant_37512" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2335823440" size="10240" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7087" name="__module.model.gpt_neox.layers.29.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
+		<layer id="4572" name="__module.model.gpt_neox.layers.28.post_attention_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -108480,37 +69525,17 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7088" name="Constant_299360_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2417818834" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7089" name="Constant_299360" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="4573" name="Constant_37513" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2335833680" size="10240" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7090" name="__module.model.gpt_neox.layers.29.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+		<layer id="4574" name="__module.model.gpt_neox.layers.28.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -108525,15 +69550,15 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4762">
+				<port id="2" precision="FP32" names="3609">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7091" name="self.model.gpt_neox.layers.29.mlp.dense_h_to_4h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 2560" offset="2417823954" size="26214400" />
+		<layer id="4575" name="self.model.gpt_neox.layers.28.mlp.dense_h_to_4h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 2560" offset="2335843920" size="26214400" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>10240</dim>
@@ -108541,7 +69566,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7092" name="Convert_22604" type="Convert" version="opset1">
+		<layer id="4576" name="Convert_74744" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -108556,8 +69581,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7093" name="self.model.gpt_neox.layers.29.mlp.dense_h_to_4h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 1" offset="2444038354" size="10240" />
+		<layer id="4577" name="self.model.gpt_neox.layers.28.mlp.dense_h_to_4h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 1" offset="2362058320" size="10240" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>10240</dim>
@@ -108565,7 +69590,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7094" name="Convert_22607" type="Convert" version="opset1">
+		<layer id="4578" name="Convert_74747" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -108580,7 +69605,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7095" name="self.model.gpt_neox.layers.29.mlp.dense_h_to_4h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="4579" name="self.model.gpt_neox.layers.28.mlp.dense_h_to_4h.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -108599,8 +69624,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7096" name="self.model.gpt_neox.layers.29.mlp.dense_h_to_4h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="10240, 1" offset="2444048594" size="20480" />
+		<layer id="4580" name="self.model.gpt_neox.layers.28.mlp.dense_h_to_4h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="10240, 1" offset="2362068560" size="20480" />
 			<output>
 				<port id="0" precision="FP16">
 					<dim>10240</dim>
@@ -108608,7 +69633,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7097" name="self.model.gpt_neox.layers.29.mlp.dense_h_to_4h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="4581" name="self.model.gpt_neox.layers.28.mlp.dense_h_to_4h.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -108627,7 +69652,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7098" name="self.model.gpt_neox.layers.29.mlp.dense_h_to_4h.weight" type="Convert" version="opset1">
+		<layer id="4582" name="__module.model.gpt_neox.layers.28.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
@@ -108645,7 +69670,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7099" name="__module.model.gpt_neox.layers.29.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="4583" name="__module.model.gpt_neox.layers.28.mlp.dense_h_to_4h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -108666,37 +69691,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7100" name="Constant_299361_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 10240" offset="2444069074" size="20480" />
+		<layer id="4584" name="self.model.gpt_neox.layers.28.mlp.dense_h_to_4h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="10240" offset="2362089040" size="20480" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.28.mlp.dense_h_to_4h.bias">
 					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7101" name="Constant_299361" type="Convert" version="opset1">
+		<layer id="4585" name="__module.model.gpt_neox.layers.28.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
 			</rt_info>
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16">
 					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7102" name="__module.model.gpt_neox.layers.29.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1">
+		<layer id="4586" name="__module.model.gpt_neox.layers.28.mlp.dense_h_to_4h/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -108705,20 +69724,18 @@
 					<dim>10240</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4767">
+				<port id="2" precision="FP32" names="3614">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7103" name="__module.model.gpt_neox.layers.29.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
+		<layer id="4587" name="__module.model.gpt_neox.layers.28.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
 			<data approximation_mode="ERF" />
 			<input>
 				<port id="0" precision="FP32">
@@ -108728,15 +69745,15 @@
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="4768">
+				<port id="1" precision="FP32" names="3615">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7104" name="self.model.gpt_neox.layers.29.mlp.dense_4h_to_h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 10240" offset="2444089554" size="26214400" />
+		<layer id="4588" name="self.model.gpt_neox.layers.28.mlp.dense_4h_to_h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 10240" offset="2362109520" size="26214400" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>2560</dim>
@@ -108744,7 +69761,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7105" name="Convert_22614" type="Convert" version="opset1">
+		<layer id="4589" name="Convert_74754" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -108759,8 +69776,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7106" name="self.model.gpt_neox.layers.29.mlp.dense_4h_to_h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="2470303954" size="2560" />
+		<layer id="4590" name="self.model.gpt_neox.layers.28.mlp.dense_4h_to_h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="2388323920" size="2560" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>2560</dim>
@@ -108768,7 +69785,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7107" name="Convert_22617" type="Convert" version="opset1">
+		<layer id="4591" name="Convert_74757" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -108783,7 +69800,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7108" name="self.model.gpt_neox.layers.29.mlp.dense_4h_to_h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="4592" name="self.model.gpt_neox.layers.28.mlp.dense_4h_to_h.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -108802,8 +69819,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7109" name="self.model.gpt_neox.layers.29.mlp.dense_4h_to_h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="2470306514" size="5120" />
+		<layer id="4593" name="self.model.gpt_neox.layers.28.mlp.dense_4h_to_h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="2388326480" size="5120" />
 			<output>
 				<port id="0" precision="FP16">
 					<dim>2560</dim>
@@ -108811,7 +69828,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7110" name="self.model.gpt_neox.layers.29.mlp.dense_4h_to_h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="4594" name="self.model.gpt_neox.layers.28.mlp.dense_4h_to_h.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -108830,7 +69847,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7111" name="self.model.gpt_neox.layers.29.mlp.dense_4h_to_h.weight" type="Convert" version="opset1">
+		<layer id="4595" name="__module.model.gpt_neox.layers.28.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
@@ -108848,7 +69865,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7112" name="__module.model.gpt_neox.layers.29.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="4596" name="__module.model.gpt_neox.layers.28.mlp.dense_4h_to_h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -108869,37 +69886,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7113" name="Constant_299362_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2470311634" size="5120" />
+		<layer id="4597" name="self.model.gpt_neox.layers.28.mlp.dense_4h_to_h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="2560" offset="2388331600" size="5120" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.28.mlp.dense_4h_to_h.bias">
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7114" name="Constant_299362" type="Convert" version="opset1">
+		<layer id="4598" name="__module.model.gpt_neox.layers.28.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
 			</rt_info>
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16">
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7115" name="__module.model.gpt_neox.layers.29.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1">
+		<layer id="4599" name="__module.model.gpt_neox.layers.28.mlp.dense_4h_to_h/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -108908,50 +69919,28 @@
 					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4771,input.121">
+				<port id="2" precision="FP32" names="3618,input.115">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7116" name="Constant_299363_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2470316754" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7117" name="Constant_299363" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="4600" name="Constant_37514" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2388336720" size="10240" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7118" name="__module.model.gpt_neox.layers.29.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
+		<layer id="4601" name="__module.model.gpt_neox.layers.28.input_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -108973,37 +69962,17 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7119" name="Constant_299364_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2470321874" size="5120" />
+		<layer id="4602" name="Constant_37515" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2388346960" size="10240" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7120" name="Constant_299364" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7121" name="__module.model.gpt_neox.layers.29.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+		<layer id="4603" name="__module.model.gpt_neox.layers.28.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -109018,15 +69987,15 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4637">
+				<port id="2" precision="FP32" names="3521,hidden_states.113">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7122" name="self.model.gpt_neox.layers.29.attention.query_key_value.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 2560" offset="2470326994" size="19660800" />
+		<layer id="4604" name="self.model.gpt_neox.layers.28.attention.query_key_value.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 2560" offset="2388357200" size="19660800" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>7680</dim>
@@ -109034,7 +70003,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7123" name="Convert_22584" type="Convert" version="opset1">
+		<layer id="4605" name="Convert_75132" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -109049,8 +70018,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7124" name="self.model.gpt_neox.layers.29.attention.query_key_value.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 1" offset="2489987794" size="7680" />
+		<layer id="4606" name="self.model.gpt_neox.layers.28.attention.query_key_value.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 1" offset="2408018000" size="7680" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>7680</dim>
@@ -109058,7 +70027,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7125" name="Convert_22587" type="Convert" version="opset1">
+		<layer id="4607" name="Convert_75135" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -109073,7 +70042,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7126" name="self.model.gpt_neox.layers.29.attention.query_key_value.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="4608" name="self.model.gpt_neox.layers.28.attention.query_key_value.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -109092,8 +70061,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7127" name="self.model.gpt_neox.layers.29.attention.query_key_value.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="7680, 1" offset="2489995474" size="15360" />
+		<layer id="4609" name="self.model.gpt_neox.layers.28.attention.query_key_value.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="7680, 1" offset="2408025680" size="15360" />
 			<output>
 				<port id="0" precision="FP16">
 					<dim>7680</dim>
@@ -109101,7 +70070,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7128" name="self.model.gpt_neox.layers.29.attention.query_key_value.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="4610" name="self.model.gpt_neox.layers.28.attention.query_key_value.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -109120,11 +70089,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7129" name="self.model.gpt_neox.layers.29.attention.query_key_value.weight" type="Convert" version="opset1">
+		<layer id="4611" name="self.model.gpt_neox.layers.28.attention.query_key_value.weight/fq_weights_1/convert" type="Convert" version="opset1">
 			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
 			<input>
 				<port id="0" precision="FP16">
 					<dim>7680</dim>
@@ -109138,7 +70104,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7130" name="__module.model.gpt_neox.layers.29.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="4612" name="__module.model.gpt_neox.layers.28.attention.query_key_value/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -109159,37 +70125,17 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7131" name="Constant_299365_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 7680" offset="2490010834" size="15360" />
+		<layer id="4613" name="Constant_37516" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 7680" offset="2408041040" size="30720" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7132" name="Constant_299365" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7133" name="__module.model.gpt_neox.layers.29.attention.query_key_value/aten::linear/Add" type="Add" version="opset1">
+		<layer id="4614" name="__module.model.gpt_neox.layers.28.attention.query_key_value/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -109204,25 +70150,22 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4644,qkv.117">
+				<port id="2" precision="FP32" names="3528,qkv.113">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7134" name="__module.model.gpt_neox.layers.29.attention/prim::ListConstruct/Concat" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080154" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4615" name="__module.model.gpt_neox.layers.28.attention/prim::ListConstruct" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201115992" size="32" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7135" name="__module.model.gpt_neox.layers.29.attention/aten::view/Reshape" type="Reshape" version="opset1">
+		<layer id="4616" name="__module.model.gpt_neox.layers.28.attention/aten::view/Reshape" type="Reshape" version="opset1">
 			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -109235,7 +70178,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4648,qkv.119">
+				<port id="2" precision="FP32" names="3532,qkv.115">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -109243,41 +70186,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7136" name="Constant_279904" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4617" name="Constant_18356" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7137" name="Constant_279907" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4618" name="Constant_18358" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7138" name="Constant_279910" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4619" name="Constant_18360" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7139" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4620" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Slice" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -109286,17 +70219,20 @@
 					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4649">
+				<port id="5" precision="FP32" names="3533">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -109304,15 +70240,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7140" name="Constant_17072" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="4621" name="Constant_18369" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64" names="4650">
+				<port id="0" precision="I64" names="3534">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7141" name="__module.model.gpt_neox.layers.29.attention/aten::permute/Transpose" type="Transpose" version="opset1">
+		<layer id="4622" name="__module.model.gpt_neox.layers.28.attention/aten::permute/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -109325,7 +70261,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4651,query.175">
+				<port id="2" precision="FP32" names="3535,query.57">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -109333,41 +70269,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7142" name="Constant_279916" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4623" name="Constant_18410" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7143" name="Constant_279919" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4624" name="Constant_18412" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7144" name="Constant_279922" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4625" name="Constant_18414" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7145" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4626" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Slice_1" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -109376,17 +70302,20 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4658,q.59">
+				<port id="5" precision="FP32" names="3542,q.57">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -109394,219 +70323,152 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7146" name="self.model.gpt_neox.layers.29.attention.rotary_emb.cos_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201080346" size="81920" />
-			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.29.attention.rotary_emb.cos_cached">
-					<dim>2048</dim>
+		<layer id="4627" name="__module.model.gpt_neox.layers.28.attention/aten::mul/Multiply" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="7147" name="self.model.gpt_neox.layers.29.attention.rotary_emb.cos_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
+				<port id="2" precision="FP32" names="3548">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7148" name="__module.model.gpt_neox.layers.29.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4628" name="Constant_18568" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7149" name="Constant_279928" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4629" name="Constant_18570" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7150" name="Constant_279931" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7151" name="Constant_279934" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4630" name="Constant_18572" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7152" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4631" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Slice_2" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>240</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4652">
-					<dim>-1</dim>
+				<port id="5" precision="FP32" names="3558,x2.113">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7153" name="Constant_17086" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="4632" name="Constant_37517" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="0" precision="I64" names="4653">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7154" name="__module.model.gpt_neox.layers.29.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
+		<layer id="4633" name="__module.model.gpt_neox.layers.28.attention/aten::neg/Multiply" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4654,key.175">
+				<port id="2" precision="FP32" names="3559">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7155" name="__module.model.gpt_neox.layers.29.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7156" name="__module.model.gpt_neox.layers.29.attention/aten::size/Constant_2" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="7157" name="__module.model.gpt_neox.layers.29.attention/aten::size/Gather_2" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>10</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="4662" />
-			</output>
-		</layer>
-		<layer id="7158" name="Constant_338183_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="7159" name="Constant_338183" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
 			</output>
 		</layer>
-		<layer id="7160" name="Constant_338176" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
+		<layer id="4634" name="Constant_18465" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7161" name="Constant_338178" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
+		<layer id="4635" name="Constant_18509" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7162" name="Constant_338180" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
+		<layer id="4636" name="Constant_18511" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7163" name="Concat_338181" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="4637" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Slice_3" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -109617,163 +70479,125 @@
 				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
 				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7164" name="Broadcast_338184" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="5" precision="FP32" names="3553">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7165" name="ReadValue_330513" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.29.keypresent.29.key" variable_type="f32" variable_shape="?,32,?,80" />
+		<layer id="4638" name="__module.model.gpt_neox.layers.28.attention/aten::cat/Concat" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="past_key_values.29.key">
+				<port id="2" precision="FP32" names="3561">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7166" name="Constant_323559" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64" />
-			</output>
-		</layer>
-		<layer id="7167" name="Gather_323560" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="4639" name="__module.model.gpt_neox.layers.28.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="I32">
+				<port id="1" precision="FP32">
 					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
-				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="3" precision="FP32">
+				<port id="2" precision="FP32" names="3562">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7168" name="__module.model.gpt_neox.layers.29.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="4640" name="__module.model.gpt_neox.layers.28.attention/aten::add/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="3563">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7169" name="__module.model.gpt_neox.layers.29.attention/aten::size/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4641" name="Constant_18420" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="7170" name="__module.model.gpt_neox.layers.29.attention/aten::size/Gather_3" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="4664" />
-			</output>
-		</layer>
-		<layer id="7171" name="__module.model.gpt_neox.layers.29.attention/aten::add_/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" names="4663,4667,4668,seq_len.117" />
-			</output>
-		</layer>
-		<layer id="7172" name="__module.model.gpt_neox.layers.29.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
 			</output>
 		</layer>
-		<layer id="7173" name="__module.model.gpt_neox.layers.29.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="4642" name="Constant_18422" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7174" name="__module.model.gpt_neox.layers.29.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4643" name="Constant_18424" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7175" name="__module.model.gpt_neox.layers.29.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4644" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Slice_4" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -109784,197 +70608,192 @@
 				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4671,4674,cos.117">
+				<port id="5" precision="FP32" names="3543">
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7176" name="Constant_38371" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="7177" name="__module.model.gpt_neox.layers.29.attention/aten::index/Gather" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="4645" name="__module.model.gpt_neox.layers.28.attention/aten::cat/Concat_1" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
-				<port id="1" precision="I32">
+				<port id="1" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="4677">
+				<port id="2" precision="FP32" names="3581">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7178" name="__module.model.gpt_neox.layers.29.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64" />
-			</input>
+		<layer id="4646" name="Constant_53223" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="2" precision="FP32" names="4678,cos.119">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
+				<port id="0" precision="FP32" />
 			</output>
 		</layer>
-		<layer id="7179" name="__module.model.gpt_neox.layers.29.attention/aten::mul/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
+		<layer id="4647" name="Constant_53216" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
-				<port id="2" precision="FP32" names="4682">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7180" name="Constant_279939" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4648" name="Constant_53218" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7181" name="Constant_279938" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4649" name="Constant_53220" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7182" name="__module.model.gpt_neox.layers.29.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="4650" name="Concat_53221" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
+				<port id="4" precision="I64">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7183" name="__module.model.gpt_neox.layers.29.attention/aten::size/Constant_5" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="7184" name="__module.model.gpt_neox.layers.29.attention/aten::size/Gather_5" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="4651" name="Broadcast_53224" type="Broadcast" version="opset3">
+			<data mode="numpy" />
 			<input>
-				<port id="0" precision="I64">
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="4683,4688" />
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="7185" name="__module.model.gpt_neox.layers.29.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
+		<layer id="4652" name="ReadValue_50267" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.28.keypresent.28.key" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64" />
+				<port id="1" precision="FP32" names="past_key_values.28.key">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4653" name="Constant_47946" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="7186" name="__module.model.gpt_neox.layers.29.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1">
+		<layer id="4654" name="Gather_47947" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
-				<port id="0" precision="I64" />
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I32">
+					<dim>-1</dim>
+				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="1" precision="I64" names="4685,4686,4690,4691" />
+				<port id="3" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="7187" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Constant_1" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4655" name="Constant_18374" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="7188" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="4656" name="Constant_18376" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7189" name="Constant_279937" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4657" name="Constant_18378" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7190" name="ScatterUpdate_279940" type="ScatterUpdate" version="opset3">
+		<layer id="4658" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Slice_5" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -109982,148 +70801,155 @@
 				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
 				<port id="4" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="7191" name="Constant_279943" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="5" precision="FP32" names="3536">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7192" name="Constant_279946" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4659" name="Constant_18387" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="I64" names="3537">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7193" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4660" name="__module.model.gpt_neox.layers.28.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4692,x2.117">
+				<port id="2" precision="FP32" names="3538,key.113">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7194" name="Constant_299366_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
+		<layer id="4661" name="Constant_18430" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4662" name="Constant_18432" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4663" name="Constant_18434" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7195" name="Constant_299366" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="4664" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Slice_6" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="5" precision="FP32" names="3544,k.57">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7196" name="__module.model.gpt_neox.layers.29.attention/aten::neg/Multiply" type="Multiply" version="opset1">
+		<layer id="4665" name="__module.model.gpt_neox.layers.28.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
+					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4693">
+				<port id="2" precision="FP32" names="3564">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7197" name="Constant_279991" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4666" name="Constant_18711" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7198" name="Constant_279996" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4667" name="Constant_18713" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4668" name="Constant_18715" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7199" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4669" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Slice_7" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -110132,17 +70958,20 @@
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4687">
+				<port id="5" precision="FP32" names="3574,x2.115">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -110150,8 +70979,19 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7200" name="__module.model.gpt_neox.layers.29.attention/aten::cat/Concat" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="4670" name="Constant_37518" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4671" name="__module.model.gpt_neox.layers.28.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -110160,75 +71000,51 @@
 					<dim>10</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4695">
+				<port id="2" precision="FP32" names="3575">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7201" name="self.model.gpt_neox.layers.29.attention.rotary_emb.sin_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201162356" size="81920" />
-			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.29.attention.rotary_emb.sin_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7202" name="self.model.gpt_neox.layers.29.attention.rotary_emb.sin_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</input>
+		<layer id="4672" name="Constant_18608" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7203" name="__module.model.gpt_neox.layers.29.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4673" name="Constant_18652" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7204" name="__module.model.gpt_neox.layers.29.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4674" name="Constant_18654" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7205" name="__module.model.gpt_neox.layers.29.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4675" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Slice_8" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>2048</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
@@ -110240,60 +71056,45 @@
 				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4672,4675,sin.117">
+				<port id="5" precision="FP32" names="3569">
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7206" name="Constant_38374" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="7207" name="__module.model.gpt_neox.layers.29.attention/aten::index/Gather_1" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="4676" name="__module.model.gpt_neox.layers.28.attention/aten::cat/Concat_2" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="4680">
-					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>10</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="7208" name="__module.model.gpt_neox.layers.29.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
+				<port id="1" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>10</dim>
 				</port>
-				<port id="1" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4681,sin.119">
+				<port id="2" precision="FP32" names="3577">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7209" name="__module.model.gpt_neox.layers.29.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
+		<layer id="4677" name="__module.model.gpt_neox.layers.28.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -110310,7 +71111,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4696">
+				<port id="2" precision="FP32" names="3578">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -110318,7 +71119,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7210" name="__module.model.gpt_neox.layers.29.attention/aten::add/Add" type="Add" version="opset1">
+		<layer id="4678" name="__module.model.gpt_neox.layers.28.attention/aten::add/Add_1" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -110335,7 +71136,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4697">
+				<port id="2" precision="FP32" names="3579">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -110343,41 +71144,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7211" name="Constant_280042" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4679" name="Constant_18440" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7212" name="Constant_280045" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4680" name="Constant_18442" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7213" name="Constant_280048" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4681" name="Constant_18444" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7214" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4682" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Slice_9" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -110386,17 +71177,20 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4659">
+				<port id="5" precision="FP32" names="3545">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -110404,7 +71198,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7215" name="__module.model.gpt_neox.layers.29.attention/aten::cat/Concat_1" type="Concat" version="opset1">
+		<layer id="4683" name="__module.model.gpt_neox.layers.28.attention/aten::cat/Concat_3" type="Concat" version="opset1">
 			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
@@ -110421,7 +71215,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4715,4725,query.177,query.179">
+				<port id="2" precision="FP32" names="3583">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -110429,41 +71223,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7216" name="Constant_280054" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7217" name="Constant_280057" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7218" name="Constant_280060" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7219" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4684" name="__module.model.gpt_neox.layers.28.attention/aten::cat/Concat_4" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -110471,163 +71232,57 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="4660,k.59">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7220" name="__module.model.gpt_neox.layers.29.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4698">
+				<port id="2" precision="FP32" names="present.28.key">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7221" name="Constant_280064" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4685" name="Constant_53235" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
+				<port id="0" precision="FP32" />
 			</output>
 		</layer>
-		<layer id="7222" name="Constant_280063" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4686" name="Constant_53228" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7223" name="__module.model.gpt_neox.layers.29.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7224" name="__module.model.gpt_neox.layers.29.attention/aten::size/Constant_7" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4687" name="Constant_53230" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="7225" name="__module.model.gpt_neox.layers.29.attention/aten::size/Gather_7" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="4699,4704" />
-			</output>
-		</layer>
-		<layer id="7226" name="__module.model.gpt_neox.layers.29.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" />
-			</output>
-		</layer>
-		<layer id="7227" name="__module.model.gpt_neox.layers.29.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-			</input>
-			<output>
-				<port id="1" precision="I64" names="4701,4702,4706,4707" />
-			</output>
-		</layer>
-		<layer id="7228" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="7229" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7230" name="Constant_280062" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4688" name="Constant_53232" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7231" name="ScatterUpdate_280065" type="ScatterUpdate" version="opset3">
+		<layer id="4689" name="Concat_53233" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -110635,7 +71290,7 @@
 				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
@@ -110645,318 +71300,318 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7232" name="Constant_280068" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
+		<layer id="4690" name="Broadcast_53236" type="Broadcast" version="opset3">
+			<data mode="numpy" />
+			<input>
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="7233" name="Constant_280071" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7234" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4691" name="ReadValue_50269" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.28.valuepresent.28.value" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4708,x2.119">
+				<port id="1" precision="FP32" names="past_key_values.28.value">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7235" name="Constant_299367_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7236" name="Constant_299367" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
+		<layer id="4692" name="Constant_47949" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="7237" name="__module.model.gpt_neox.layers.29.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4693" name="Gather_47950" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="1" precision="I32">
+					<dim>-1</dim>
 				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4709">
+				<port id="3" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7238" name="Constant_280116" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4694" name="Constant_18392" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7239" name="Constant_280121" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4695" name="Constant_18394" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4696" name="Constant_18396" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7240" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4697" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Slice_10" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4703">
+				<port id="5" precision="FP32" names="3539">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7241" name="__module.model.gpt_neox.layers.29.attention/aten::cat/Concat_2" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="4698" name="Constant_18405" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
+			<output>
+				<port id="0" precision="I64" names="3540">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4699" name="__module.model.gpt_neox.layers.28.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-				<port id="1" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4711">
+				<port id="2" precision="FP32" names="3541">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7242" name="__module.model.gpt_neox.layers.29.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4700" name="__module.model.gpt_neox.layers.28.attention/aten::cat/Concat_5" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4712">
+				<port id="2" precision="FP32" names="present.28.value">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7243" name="__module.model.gpt_neox.layers.29.attention/aten::add/Add_1" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4701" name="Constant_18787" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4702" name="ShapeOf_40700" type="ShapeOf" version="opset3">
+			<data output_type="i64" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4713">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7244" name="Constant_280166" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4703" name="Constant_40701" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116168" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7245" name="Constant_280169" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4704" name="Constant_40702" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
+				<port id="0" precision="I64" />
+			</output>
+		</layer>
+		<layer id="4705" name="Gather_40703" type="Gather" version="opset8">
+			<data batch_dims="0" />
+			<input>
 				<port id="0" precision="I64">
 					<dim>4</dim>
 				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64" />
+			</input>
+			<output>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="7246" name="Constant_280172" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4706" name="Add_40706" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="I64" names="3588">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4707" name="Constant_18790" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7247" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4708" name="__module.model.gpt_neox.layers.28.attention/aten::slice/Slice_14" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4661">
+				<port id="5" precision="FP32" names="3592">
+					<dim>-1</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>60</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7248" name="__module.model.gpt_neox.layers.29.attention/aten::cat/Concat_3" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="4709" name="__module.model.gpt_neox.layers.28.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+			<data causal="false" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>60</dim>
+					<dim>80</dim>
+				</port>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="3" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4717">
+				<port id="4" precision="FP32" names="3593,attn_output.169">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -110964,8 +71619,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7249" name="__module.model.gpt_neox.layers.29.attention/aten::to/Convert_1" type="Concat" version="opset1">
-			<data axis="-2" />
+		<layer id="4710" name="__module.model.gpt_neox.layers.28.attention/aten::transpose/Constant" type="Const" version="opset1">
+			<data element_type="i32" shape="4" offset="201116200" size="16" />
+			<output>
+				<port id="0" precision="I32">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4711" name="__module.model.gpt_neox.layers.28.attention/aten::transpose/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -110973,877 +71635,1023 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+				<port id="1" precision="I32">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="present.29.key">
+				<port id="2" precision="FP32" names="3594">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7250" name="Constant_338196_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
+		<layer id="4712" name="Constant_41012" type="Const" version="opset1">
+			<data element_type="i64" shape="3" offset="201116216" size="24" />
 			<output>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="I64">
+					<dim>3</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="7251" name="Constant_338196" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="4713" name="__module.model.gpt_neox.layers.28.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
+			<data special_zero="true" />
 			<input>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>3</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" />
+				<port id="2" precision="FP32" names="3597">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="7252" name="Constant_338189" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
+		<layer id="4714" name="self.model.gpt_neox.layers.28.attention.dense.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 2560" offset="2408071760" size="6553600" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7253" name="Constant_338191" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
+		<layer id="4715" name="Convert_75484" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7254" name="Constant_338193" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
+		<layer id="4716" name="self.model.gpt_neox.layers.28.attention.dense.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="2414625360" size="2560" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="U8">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7255" name="Concat_338194" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="4717" name="Convert_75487" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64">
+				<port id="0" precision="U8">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64">
+			</input>
+			<output>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+			</output>
+		</layer>
+		<layer id="4718" name="self.model.gpt_neox.layers.28.attention.dense.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7256" name="Broadcast_338197" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
+		<layer id="4719" name="self.model.gpt_neox.layers.28.attention.dense.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="2414627920" size="5120" />
 			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7257" name="ReadValue_330515" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.29.valuepresent.29.value" variable_type="f32" variable_shape="?,32,?,80" />
+		<layer id="4720" name="self.model.gpt_neox.layers.28.attention.dense.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="past_key_values.29.value">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7258" name="Constant_323562" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
+		<layer id="4721" name="self.model.gpt_neox.layers.28.attention.dense.weight/fq_weights_1/convert" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64" />
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="7259" name="Gather_323563" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="4722" name="__module.model.gpt_neox.layers.28.attention.dense/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="3" precision="FP32">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7260" name="Constant_280178" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7261" name="Constant_280181" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7262" name="Constant_280184" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4723" name="Constant_37519" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2414633040" size="10240" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7263" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4724" name="__module.model.gpt_neox.layers.28.attention.dense/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4655">
+				<port id="2" precision="FP32" names="3600,3602,input.113">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7264" name="Constant_17100" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="4656">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7265" name="__module.model.gpt_neox.layers.29.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
+		<layer id="4725" name="__module.model.gpt_neox.layers.28/aten::add/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4657">
+				<port id="2" precision="FP32" names="3620">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7266" name="__module.model.gpt_neox.layers.29.attention/aten::cat/Concat_5" type="Concat" version="opset1">
-			<data axis="-2" />
+		<layer id="4726" name="__module.model.gpt_neox.layers.28/aten::add/Add_1" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="present.29.value">
+				<port id="2" precision="FP32" names="3621,3623,hidden_states.119">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7267" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Slice_11" type="Const" version="opset1">
-			<data element_type="boolean" shape="1, 1, 2048, 2048" offset="201244276" size="4194304" />
+		<layer id="4727" name="__module.model.gpt_neox.layers.29.input_layernorm/aten::layer_norm/Multiply" type="Const" version="opset1">
+			<data element_type="i32" shape="1" offset="128867668" size="4" />
 			<output>
-				<port id="0" precision="BOOL" names="4734">
-					<dim>1</dim>
+				<port id="0" precision="I32">
 					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7268" name="Constant_280188" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4728" name="__module.model.gpt_neox.layers.29.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
+			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="I32">
+					<dim>1</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7269" name="Constant_280187" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162298" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4729" name="Constant_37520" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2414643280" size="10240" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="FP32">
+					<dim>1</dim>
 					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7270" name="__module.model.gpt_neox.layers.29.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="4730" name="__module.model.gpt_neox.layers.29.post_attention_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7271" name="__module.model.gpt_neox.layers.29.attention/aten::size/Constant_8" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4731" name="Constant_37521" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2414653520" size="10240" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="7272" name="__module.model.gpt_neox.layers.29.attention/aten::size/Gather_8" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="4732" name="__module.model.gpt_neox.layers.29.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="4729" />
+				<port id="2" precision="FP32" names="3721">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4733" name="self.model.gpt_neox.layers.29.mlp.dense_h_to_4h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 2560" offset="2414663760" size="26214400" />
+			<output>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="7273" name="__module.model.gpt_neox.layers.29.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="4734" name="Convert_74764" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7274" name="__module.model.gpt_neox.layers.29.attention/aten::size/Constant_9" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4735" name="self.model.gpt_neox.layers.29.mlp.dense_h_to_4h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 1" offset="2440878160" size="10240" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="7275" name="__module.model.gpt_neox.layers.29.attention/aten::size/Gather_9" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="4736" name="Convert_74767" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="4727" />
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="7276" name="__module.model.gpt_neox.layers.29.attention/aten::sub/Subtract" type="Subtract" version="opset1">
+		<layer id="4737" name="self.model.gpt_neox.layers.29.mlp.dense_h_to_4h.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64" names="4731,4732" />
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="7277" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Constant_4" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4738" name="self.model.gpt_neox.layers.29.mlp.dense_h_to_4h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="10240, 1" offset="2440888400" size="20480" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="7278" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1">
+		<layer id="4739" name="self.model.gpt_neox.layers.29.mlp.dense_h_to_4h.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7279" name="Constant_280186" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
+		<layer id="4740" name="__module.model.gpt_neox.layers.29.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<rt_info>
-				<attribute name="precise" version="0" />
+				<attribute name="decompression" version="0" />
 			</rt_info>
+			<input>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7280" name="ScatterUpdate_280189" type="ScatterUpdate" version="opset3">
+		<layer id="4741" name="__module.model.gpt_neox.layers.29.mlp.dense_h_to_4h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7281" name="Constant_280190" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4742" name="self.model.gpt_neox.layers.29.mlp.dense_h_to_4h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="10240" offset="2440908880" size="20480" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.29.mlp.dense_h_to_4h.bias">
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7282" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1">
+		<layer id="4743" name="__module.model.gpt_neox.layers.29.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<rt_info>
+				<attribute name="decompression" version="0" />
+			</rt_info>
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
+				<port id="0" precision="BF16">
+					<dim>10240</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7283" name="ScatterUpdate_280191" type="ScatterUpdate" version="opset3">
+		<layer id="4744" name="__module.model.gpt_neox.layers.29.mlp.dense_h_to_4h/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7284" name="Constant_280194" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438612" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="2" precision="FP32" names="3726">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7285" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4745" name="__module.model.gpt_neox.layers.29.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
+			<data approximation_mode="ERF" />
 			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>3</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="BOOL" names="4735">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="1" precision="FP32" names="3727">
 					<dim>-1</dim>
-					<dim>2048</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7286" name="Constant_280267" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4746" name="self.model.gpt_neox.layers.29.mlp.dense_4h_to_h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 10240" offset="2440929360" size="26214400" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7287" name="Constant_280268" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4747" name="Convert_74774" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7288" name="Constant_280264" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4748" name="self.model.gpt_neox.layers.29.mlp.dense_4h_to_h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="2467143760" size="2560" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="U8">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7289" name="Constant_280263" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4749" name="Convert_74777" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I32">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7290" name="ScatterUpdate_280269" type="ScatterUpdate" version="opset3">
+		<layer id="4750" name="self.model.gpt_neox.layers.29.mlp.dense_4h_to_h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4751" name="self.model.gpt_neox.layers.29.mlp.dense_4h_to_h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="2467146320" size="5120" />
+			<output>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+			</output>
+		</layer>
+		<layer id="4752" name="self.model.gpt_neox.layers.29.mlp.dense_4h_to_h.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7291" name="Constant_280272" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
+		<layer id="4753" name="__module.model.gpt_neox.layers.29.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<rt_info>
-				<attribute name="precise" version="0" />
+				<attribute name="decompression" version="0" />
 			</rt_info>
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7292" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4754" name="__module.model.gpt_neox.layers.29.mlp.dense_4h_to_h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="BOOL" names="4736,4737">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7293" name="__module.model.gpt_neox.layers.29.attention/aten::where/Convert_1_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
+		<layer id="4755" name="self.model.gpt_neox.layers.29.mlp.dense_4h_to_h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="2560" offset="2467151440" size="5120" />
 			<output>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.29.mlp.dense_4h_to_h.bias">
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="7294" name="__module.model.gpt_neox.layers.29.attention/aten::where/Convert_1" type="Convert" version="opset1">
+		<layer id="4756" name="__module.model.gpt_neox.layers.29.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
 			</rt_info>
 			<input>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="BF16">
+					<dim>2560</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="7295" name="__module.model.gpt_neox.layers.29.attention/aten::full/Convert" type="Const" version="opset1">
-			<data element_type="f32" shape="" offset="205438636" size="4" />
-			<output>
-				<port id="0" precision="FP32" names="4724" />
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="7296" name="__module.model.gpt_neox.layers.29.attention/aten::where/Select" type="Select" version="opset1">
+		<layer id="4757" name="__module.model.gpt_neox.layers.29.mlp.dense_4h_to_h/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="FP32" />
-				<port id="2" precision="FP32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="4738,causal_mask.117">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="3730,input.119">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7297" name="Add_290512" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
+		<layer id="4758" name="Constant_37522" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2467156560" size="10240" />
+			<output>
 				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4759" name="__module.model.gpt_neox.layers.29.input_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7298" name="ShapeOf_290513" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="4760" name="Constant_37523" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2467166800" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4761" name="__module.model.gpt_neox.layers.29.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
 					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="3633,hidden_states.117">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="4762" name="self.model.gpt_neox.layers.29.attention.query_key_value.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 2560" offset="2467177040" size="19660800" />
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7299" name="Constant_254764" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4763" name="Convert_75143" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7300" name="__module.model.gpt_neox.layers.29.attention/aten::size/Constant_10" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4764" name="self.model.gpt_neox.layers.29.attention.query_key_value.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 1" offset="2486837840" size="7680" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="7301" name="__module.model.gpt_neox.layers.29.attention/aten::size/Gather_10" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="4765" name="Convert_75146" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
+				<port id="0" precision="U8">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="4722">
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7302" name="__module.model.gpt_neox.layers.29.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="4766" name="self.model.gpt_neox.layers.29.attention.query_key_value.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64" names="4739">
-					<dim>4</dim>
+				<port id="2" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4767" name="self.model.gpt_neox.layers.29.attention.query_key_value.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="7680, 1" offset="2486845520" size="15360" />
+			<output>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7303" name="__module.model.gpt_neox.layers.29.attention/aten::expand/Abs" type="Abs" version="opset1">
+		<layer id="4768" name="self.model.gpt_neox.layers.29.attention.query_key_value.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7304" name="Maximum_290514" type="Maximum" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4769" name="self.model.gpt_neox.layers.29.attention.query_key_value.weight/fq_weights_1/convert" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7305" name="__module.model.gpt_neox.layers.29.attention/aten::add/Add_2" type="Broadcast" version="opset3">
-			<data mode="numpy" />
+		<layer id="4770" name="__module.model.gpt_neox.layers.29.attention.query_key_value/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4741">
-					<dim>-1</dim>
-					<dim>-1</dim>
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7306" name="__module.model.gpt_neox.layers.29.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
-			<data causal="false" />
+		<layer id="4771" name="Constant_37524" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 7680" offset="2486860880" size="30720" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>7680</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4772" name="__module.model.gpt_neox.layers.29.attention.query_key_value/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>7680</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>7680</dim>
 				</port>
-				<port id="2" precision="FP32">
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="3640,qkv.117">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>7680</dim>
 				</port>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
+			</output>
+		</layer>
+		<layer id="4773" name="__module.model.gpt_neox.layers.29.attention/prim::ListConstruct" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201115992" size="32" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4774" name="__module.model.gpt_neox.layers.29.attention/aten::view/Reshape" type="Reshape" version="opset1">
+			<data special_zero="true" />
+			<input>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>7680</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4742,4743,sdpa_result.59,tensor.117">
+				<port id="2" precision="FP32" names="3644,qkv.119">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7307" name="Constant_17541" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="4775" name="Constant_18986" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="I64" names="4744">
-					<dim>4</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4776" name="Constant_18988" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4777" name="Constant_18990" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7308" name="__module.model.gpt_neox.layers.29.attention/aten::permute/Transpose_3" type="Transpose" version="opset1">
+		<layer id="4778" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Slice" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4745">
+				<port id="5" precision="FP32" names="3645">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -111851,19 +72659,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7309" name="__module.model.gpt_neox.layers.29.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438650" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4779" name="Constant_18999" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="0" precision="I64" names="3646">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7310" name="__module.model.gpt_neox.layers.29.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
-			<data special_zero="true" />
+		<layer id="4780" name="__module.model.gpt_neox.layers.29.attention/aten::permute/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -111872,1125 +72676,1139 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>3</dim>
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4750">
+				<port id="2" precision="FP32" names="3647,query.59">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7311" name="self.model.gpt_neox.layers.29.attention.dense.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 2560" offset="2490026194" size="6553600" />
+		<layer id="4781" name="Constant_19040" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7312" name="Convert_22594" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="4782" name="Constant_19042" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7313" name="self.model.gpt_neox.layers.29.attention.dense.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="2496579794" size="2560" />
+		<layer id="4783" name="Constant_19044" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7314" name="Convert_22597" type="Convert" version="opset1">
-			<data destination_type="f16" />
+		<layer id="4784" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Slice_1" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>1</dim>
+				<port id="5" precision="FP32" names="3654,q.59">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7315" name="self.model.gpt_neox.layers.29.attention.dense.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="4785" name="__module.model.gpt_neox.layers.29.attention/aten::mul/Multiply" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="3660">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+			</output>
+		</layer>
+		<layer id="4786" name="Constant_19198" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="4787" name="Constant_19200" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7316" name="self.model.gpt_neox.layers.29.attention.dense.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="2496582354" size="5120" />
+		<layer id="4788" name="Constant_19202" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7317" name="self.model.gpt_neox.layers.29.attention.dense.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4789" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Slice_2" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="5" precision="FP32" names="3670,x2.117">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7318" name="self.model.gpt_neox.layers.29.attention.dense.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="4790" name="Constant_37525" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7319" name="__module.model.gpt_neox.layers.29.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
+		<layer id="4791" name="__module.model.gpt_neox.layers.29.attention/aten::neg/Multiply" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>2560</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="3671">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7320" name="Constant_299368_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2496587474" size="5120" />
+		<layer id="4792" name="Constant_19095" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7321" name="Constant_299368" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+		<layer id="4793" name="Constant_19139" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="4794" name="Constant_19141" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7322" name="__module.model.gpt_neox.layers.29.attention.dense/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4795" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Slice_3" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="FP32">
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4753,4755,input.119">
+				<port id="5" precision="FP32" names="3665">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7323" name="__module.model.gpt_neox.layers.29/aten::add/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4796" name="__module.model.gpt_neox.layers.29.attention/aten::cat/Concat" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4773">
+				<port id="2" precision="FP32" names="3673">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7324" name="__module.model.gpt_neox.layers.29/aten::add/Add_1" type="Add" version="opset1">
+		<layer id="4797" name="__module.model.gpt_neox.layers.29.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4774,4776,hidden_states.61">
+				<port id="2" precision="FP32" names="3674">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7325" name="__module.model.gpt_neox.layers.30.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867670" size="4" />
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7326" name="__module.model.gpt_neox.layers.30.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
-			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
+		<layer id="4798" name="__module.model.gpt_neox.layers.29.attention/aten::add/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="I32">
-					<dim>1</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="3675">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7327" name="Constant_299369_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2496592594" size="5120" />
+		<layer id="4799" name="Constant_19050" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7328" name="Constant_299369" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+		<layer id="4800" name="Constant_19052" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="4801" name="Constant_19054" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7329" name="__module.model.gpt_neox.layers.30.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4802" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Slice_4" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="1" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>2560</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="7330" name="Constant_299370_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2496597714" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="2" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="7331" name="Constant_299370" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
+				<port id="3" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+				<port id="5" precision="FP32" names="3655">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7332" name="__module.model.gpt_neox.layers.30.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4803" name="__module.model.gpt_neox.layers.29.attention/aten::cat/Concat_1" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4911">
+				<port id="2" precision="FP32" names="3693">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7333" name="self.model.gpt_neox.layers.30.mlp.dense_h_to_4h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 2560" offset="2496602834" size="26214400" />
+		<layer id="4804" name="Constant_53247" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
+				<port id="0" precision="FP32" />
 			</output>
 		</layer>
-		<layer id="7334" name="Convert_22644" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="4805" name="Constant_53240" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7335" name="self.model.gpt_neox.layers.30.mlp.dense_h_to_4h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 1" offset="2522817234" size="10240" />
+		<layer id="4806" name="Constant_53242" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7336" name="Convert_22647" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>10240</dim>
-					<dim>1</dim>
-				</port>
-			</input>
+		<layer id="4807" name="Constant_53244" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7337" name="self.model.gpt_neox.layers.30.mlp.dense_h_to_4h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4808" name="Concat_53245" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
+				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="2" precision="I64">
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="7338" name="self.model.gpt_neox.layers.30.mlp.dense_h_to_4h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="10240, 1" offset="2522827474" size="20480" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
+			</input>
+			<output>
+				<port id="4" precision="I64">
+					<dim>4</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="7339" name="self.model.gpt_neox.layers.30.mlp.dense_h_to_4h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4809" name="Broadcast_53248" type="Broadcast" version="opset3">
+			<data mode="numpy" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>10240</dim>
-					<dim>1</dim>
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7340" name="self.model.gpt_neox.layers.30.mlp.dense_h_to_4h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="4810" name="ReadValue_50271" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.29.keypresent.29.key" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="1" precision="FP32" names="past_key_values.29.key">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7341" name="__module.model.gpt_neox.layers.30.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
+		<layer id="4811" name="Constant_47952" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64" />
+			</output>
+		</layer>
+		<layer id="4812" name="Gather_47953" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>10240</dim>
-					<dim>2560</dim>
+				<port id="1" precision="I32">
+					<dim>-1</dim>
 				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="3" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7342" name="Constant_299371_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 10240" offset="2522847954" size="20480" />
+		<layer id="4813" name="Constant_19004" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7343" name="Constant_299371" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+		<layer id="4814" name="Constant_19006" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>10240</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="4815" name="Constant_19008" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7344" name="__module.model.gpt_neox.layers.30.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4816" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Slice_5" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
-				<port id="1" precision="FP32">
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4916">
+				<port id="5" precision="FP32" names="3648">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7345" name="__module.model.gpt_neox.layers.30.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
-			<data approximation_mode="ERF" />
+		<layer id="4817" name="Constant_19017" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
+			<output>
+				<port id="0" precision="I64" names="3649">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4818" name="__module.model.gpt_neox.layers.29.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="4917">
+				<port id="2" precision="FP32" names="3650,key.117">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7346" name="self.model.gpt_neox.layers.30.mlp.dense_4h_to_h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 10240" offset="2522868434" size="26214400" />
-			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7347" name="Convert_22654" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</input>
+		<layer id="4819" name="Constant_19060" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7348" name="self.model.gpt_neox.layers.30.mlp.dense_4h_to_h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="2549082834" size="2560" />
+		<layer id="4820" name="Constant_19062" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7349" name="Convert_22657" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>2560</dim>
-					<dim>1</dim>
-				</port>
-			</input>
+		<layer id="4821" name="Constant_19064" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7350" name="self.model.gpt_neox.layers.30.mlp.dense_4h_to_h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4822" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Slice_6" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7351" name="self.model.gpt_neox.layers.30.mlp.dense_4h_to_h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="2549085394" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
+				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="7352" name="self.model.gpt_neox.layers.30.mlp.dense_4h_to_h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-				<port id="1" precision="FP16">
-					<dim>2560</dim>
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7353" name="self.model.gpt_neox.layers.30.mlp.dense_4h_to_h.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
+				<port id="5" precision="FP32" names="3656,k.59">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7354" name="__module.model.gpt_neox.layers.30.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
+		<layer id="4823" name="__module.model.gpt_neox.layers.29.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10240</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>2560</dim>
-					<dim>10240</dim>
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="3676">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7355" name="Constant_299372_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2549090514" size="5120" />
+		<layer id="4824" name="Constant_19341" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7356" name="Constant_299372" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+		<layer id="4825" name="Constant_19343" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="4826" name="Constant_19345" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7357" name="__module.model.gpt_neox.layers.30.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4827" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Slice_7" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="FP32">
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4920,input.125">
+				<port id="5" precision="FP32" names="3686,x2.119">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7358" name="Constant_299373_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2549095634" size="5120" />
+		<layer id="4828" name="Constant_37526" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7359" name="Constant_299373" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7360" name="__module.model.gpt_neox.layers.30.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
+		<layer id="4829" name="__module.model.gpt_neox.layers.29.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="3687">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7361" name="Constant_299374_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2549100754" size="5120" />
+		<layer id="4830" name="Constant_19238" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="FP16">
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4831" name="Constant_19282" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4832" name="Constant_19284" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7362" name="Constant_299374" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="4833" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Slice_8" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
 				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
+				<port id="3" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>2560</dim>
+				</port>
+			</input>
+			<output>
+				<port id="5" precision="FP32" names="3681">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7363" name="__module.model.gpt_neox.layers.30.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4834" name="__module.model.gpt_neox.layers.29.attention/aten::cat/Concat_2" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>10</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4786">
+				<port id="2" precision="FP32" names="3689">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7364" name="self.model.gpt_neox.layers.30.attention.query_key_value.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 2560" offset="2549105874" size="19660800" />
+		<layer id="4835" name="__module.model.gpt_neox.layers.29.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32" names="3690">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7365" name="Convert_22624" type="Convert" version="opset1">
-			<data destination_type="f16" />
+		<layer id="4836" name="__module.model.gpt_neox.layers.29.attention/aten::add/Add_1" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32" names="3691">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7366" name="self.model.gpt_neox.layers.30.attention.query_key_value.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 1" offset="2568766674" size="7680" />
+		<layer id="4837" name="Constant_19070" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7367" name="Convert_22627" type="Convert" version="opset1">
-			<data destination_type="f16" />
-			<input>
-				<port id="0" precision="U8">
-					<dim>7680</dim>
+		<layer id="4838" name="Constant_19072" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="4839" name="Constant_19074" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7368" name="self.model.gpt_neox.layers.30.attention.query_key_value.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4840" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Slice_9" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
+				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7369" name="self.model.gpt_neox.layers.30.attention.query_key_value.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="7680, 1" offset="2568774354" size="15360" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
+				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="7370" name="self.model.gpt_neox.layers.30.attention.query_key_value.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="3" precision="I64">
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="FP16">
-					<dim>7680</dim>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="5" precision="FP32" names="3657">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7371" name="self.model.gpt_neox.layers.30.attention.query_key_value.weight" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="4841" name="__module.model.gpt_neox.layers.29.attention/aten::cat/Concat_3" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>7680</dim>
-					<dim>2560</dim>
+				<port id="2" precision="FP32" names="3695">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7372" name="__module.model.gpt_neox.layers.30.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1">
-			<data transpose_a="false" transpose_b="true" />
+		<layer id="4842" name="__module.model.gpt_neox.layers.29.attention/aten::cat/Concat_4" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>2560</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>7680</dim>
-					<dim>2560</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
+				<port id="2" precision="FP32" names="present.29.key">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7373" name="Constant_299375_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 7680" offset="2568789714" size="15360" />
+		<layer id="4843" name="Constant_53259" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32" />
+			</output>
+		</layer>
+		<layer id="4844" name="Constant_53252" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4845" name="Constant_53254" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4846" name="Constant_53256" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7374" name="Constant_299375" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="4847" name="Concat_53257" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
 					<dim>1</dim>
-					<dim>7680</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
+				<port id="4" precision="I64">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7375" name="__module.model.gpt_neox.layers.30.attention.query_key_value/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4848" name="Broadcast_53260" type="Broadcast" version="opset3">
+			<data mode="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
+					<dim>4</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
+			</output>
+		</layer>
+		<layer id="4849" name="ReadValue_50273" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.29.valuepresent.29.value" variable_type="f32" variable_shape="?,32,?,80" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4793,qkv.121">
+				<port id="1" precision="FP32" names="past_key_values.29.value">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7376" name="__module.model.gpt_neox.layers.30.attention/prim::ListConstruct/Concat" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080154" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4850" name="Constant_47955" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="7377" name="__module.model.gpt_neox.layers.30.attention/aten::view/Reshape" type="Reshape" version="opset1">
-			<data special_zero="true" />
+		<layer id="4851" name="Gather_47956" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>7680</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="I32">
+					<dim>-1</dim>
 				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4797,qkv.123">
-					<dim>-1</dim>
+				<port id="3" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>240</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7378" name="Constant_280279" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4852" name="Constant_19022" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7379" name="Constant_280282" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4853" name="Constant_19024" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7380" name="Constant_280285" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4854" name="Constant_19026" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7381" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4855" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Slice_10" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -112999,17 +73817,20 @@
 					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4798">
+				<port id="5" precision="FP32" names="3651">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -113017,15 +73838,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7382" name="Constant_17653" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="4856" name="Constant_19035" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64" names="4799">
+				<port id="0" precision="I64" names="3652">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7383" name="__module.model.gpt_neox.layers.30.attention/aten::permute/Transpose" type="Transpose" version="opset1">
+		<layer id="4857" name="__module.model.gpt_neox.layers.29.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -113038,7 +73859,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4800,query.181">
+				<port id="2" precision="FP32" names="3653">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -113046,41 +73867,41 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7384" name="Constant_280291" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+		<layer id="4858" name="__module.model.gpt_neox.layers.29.attention/aten::cat/Concat_5" type="Concat" version="opset1">
+			<data axis="-2" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="7385" name="Constant_280294" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="present.29.value">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7386" name="Constant_280297" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4859" name="Constant_19417" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7387" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4860" name="ShapeOf_40771" type="ShapeOf" version="opset3">
+			<data output_type="i64" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -113088,1053 +73909,1197 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
+			</input>
+			<output>
 				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="4861" name="Constant_40772" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116168" size="8" />
 			<output>
-				<port id="4" precision="FP32" names="4807,q.61">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7388" name="self.model.gpt_neox.layers.30.attention.rotary_emb.cos_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201080346" size="81920" />
+		<layer id="4862" name="Constant_40773" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.30.attention.rotary_emb.cos_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="7389" name="self.model.gpt_neox.layers.30.attention.rotary_emb.cos_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="4863" name="Gather_40774" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="I64">
+					<dim>4</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
 				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="3" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7390" name="__module.model.gpt_neox.layers.30.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
+		<layer id="4864" name="Add_40777" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="7391" name="Constant_280303" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="I64" names="3700">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7392" name="Constant_280306" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4865" name="Constant_19420" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7393" name="Constant_280309" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4866" name="__module.model.gpt_neox.layers.29.attention/aten::slice/Slice_14" type="Slice" version="opset8">
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="5" precision="FP32" names="3704">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7394" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4867" name="__module.model.gpt_neox.layers.29.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+			<data causal="false" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>240</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="3" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>-1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4801">
-					<dim>-1</dim>
+				<port id="4" precision="FP32" names="3705,attn_output.175">
 					<dim>-1</dim>
 					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7395" name="Constant_17667" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="4868" name="__module.model.gpt_neox.layers.29.attention/aten::transpose/Constant" type="Const" version="opset1">
+			<data element_type="i32" shape="4" offset="201116200" size="16" />
 			<output>
-				<port id="0" precision="I64" names="4802">
+				<port id="0" precision="I32">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7396" name="__module.model.gpt_neox.layers.30.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
+		<layer id="4869" name="__module.model.gpt_neox.layers.29.attention/aten::transpose/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="I32">
 					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4803,key.181">
+				<port id="2" precision="FP32" names="3706">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7397" name="__module.model.gpt_neox.layers.30.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="4870" name="Constant_41013" type="Const" version="opset1">
+			<data element_type="i64" shape="3" offset="201116216" size="24" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>3</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4871" name="__module.model.gpt_neox.layers.29.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
+			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>80</dim>
 				</port>
+				<port id="1" precision="I64">
+					<dim>3</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="3709">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7398" name="__module.model.gpt_neox.layers.30.attention/aten::size/Constant_2" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4872" name="self.model.gpt_neox.layers.29.attention.dense.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 2560" offset="2486891600" size="6553600" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="7399" name="__module.model.gpt_neox.layers.30.attention/aten::size/Gather_2" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="4873" name="Convert_75495" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="4811" />
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="7400" name="Constant_338209_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
+		<layer id="4874" name="self.model.gpt_neox.layers.29.attention.dense.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="2493445200" size="2560" />
 			<output>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="7401" name="Constant_338209" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="4875" name="Convert_75498" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP16" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="7402" name="Constant_338202" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
-			<output>
-				<port id="0" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7403" name="Constant_338204" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64">
+		<layer id="4876" name="self.model.gpt_neox.layers.29.attention.dense.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="7404" name="Constant_338206" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
+		<layer id="4877" name="self.model.gpt_neox.layers.29.attention.dense.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="2493447760" size="5120" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7405" name="Concat_338207" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="4878" name="self.model.gpt_neox.layers.29.attention.dense.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
+			</output>
+		</layer>
+		<layer id="4879" name="self.model.gpt_neox.layers.29.attention.dense.weight/fq_weights_1/convert" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7406" name="Broadcast_338210" type="Broadcast" version="opset3">
-			<data mode="numpy" />
+		<layer id="4880" name="__module.model.gpt_neox.layers.29.attention.dense/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7407" name="ReadValue_330521" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.30.keypresent.30.key" variable_type="f32" variable_shape="?,32,?,80" />
+		<layer id="4881" name="Constant_37527" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2493452880" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4882" name="__module.model.gpt_neox.layers.29.attention.dense/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="past_key_values.30.key">
+				<port id="2" precision="FP32" names="3712,3714,input.117">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7408" name="Constant_323565" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
+		<layer id="4883" name="__module.model.gpt_neox.layers.29/aten::add/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64" />
+				<port id="2" precision="FP32" names="3732">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="7409" name="Gather_323566" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="4884" name="__module.model.gpt_neox.layers.29/aten::add/Add_1" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I32">
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="3" precision="FP32">
+				<port id="2" precision="FP32" names="3733,3735,hidden_states.123">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7410" name="__module.model.gpt_neox.layers.30.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="4885" name="__module.model.gpt_neox.layers.30.input_layernorm/aten::layer_norm/Multiply" type="Const" version="opset1">
+			<data element_type="i32" shape="1" offset="128867668" size="4" />
+			<output>
+				<port id="0" precision="I32">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4886" name="__module.model.gpt_neox.layers.30.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
+			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="I32">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7411" name="__module.model.gpt_neox.layers.30.attention/aten::size/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4887" name="Constant_37528" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2493463120" size="10240" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="7412" name="__module.model.gpt_neox.layers.30.attention/aten::size/Gather_3" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="4888" name="__module.model.gpt_neox.layers.30.post_attention_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="4813" />
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4889" name="Constant_37529" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2493473360" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="7413" name="__module.model.gpt_neox.layers.30.attention/aten::add_/Add" type="Add" version="opset1">
+		<layer id="4890" name="__module.model.gpt_neox.layers.30.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64" names="4812,4816,4817,seq_len.121" />
+				<port id="2" precision="FP32" names="3833">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="7414" name="__module.model.gpt_neox.layers.30.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4891" name="self.model.gpt_neox.layers.30.mlp.dense_h_to_4h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 2560" offset="2493483600" size="26214400" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="7415" name="__module.model.gpt_neox.layers.30.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1">
+		<layer id="4892" name="Convert_74784" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
+				<port id="0" precision="U8">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7416" name="__module.model.gpt_neox.layers.30.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4893" name="self.model.gpt_neox.layers.30.mlp.dense_h_to_4h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 1" offset="2519698000" size="10240" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="U8">
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7417" name="__module.model.gpt_neox.layers.30.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4894" name="Convert_74787" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
+				<port id="0" precision="U8">
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
+			</input>
+			<output>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I64">
+			</output>
+		</layer>
+		<layer id="4895" name="self.model.gpt_neox.layers.30.mlp.dense_h_to_4h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4820,4823,cos.121">
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7418" name="Constant_38377" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
+		<layer id="4896" name="self.model.gpt_neox.layers.30.mlp.dense_h_to_4h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="10240, 1" offset="2519708240" size="20480" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="7419" name="__module.model.gpt_neox.layers.30.attention/aten::index/Gather" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="4897" name="self.model.gpt_neox.layers.30.mlp.dense_h_to_4h.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
+				<port id="1" precision="FP16">
+					<dim>10240</dim>
+					<dim>1</dim>
 				</port>
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="4826">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="2" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7420" name="__module.model.gpt_neox.layers.30.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1">
+		<layer id="4898" name="__module.model.gpt_neox.layers.30.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<rt_info>
+				<attribute name="decompression" version="0" />
+			</rt_info>
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="0" precision="FP16">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4827,cos.123">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7421" name="__module.model.gpt_neox.layers.30.attention/aten::mul/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4899" name="__module.model.gpt_neox.layers.30.mlp.dense_h_to_4h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>10240</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4831">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7422" name="Constant_280314" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4900" name="self.model.gpt_neox.layers.30.mlp.dense_h_to_4h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="10240" offset="2519728720" size="20480" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.30.mlp.dense_h_to_4h.bias">
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7423" name="Constant_280313" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
+		<layer id="4901" name="__module.model.gpt_neox.layers.30.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<rt_info>
-				<attribute name="precise" version="0" />
+				<attribute name="decompression" version="0" />
 			</rt_info>
+			<input>
+				<port id="0" precision="BF16">
+					<dim>10240</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7424" name="__module.model.gpt_neox.layers.30.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="4902" name="__module.model.gpt_neox.layers.30.mlp.dense_h_to_4h/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>10240</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="3838">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7425" name="__module.model.gpt_neox.layers.30.attention/aten::size/Constant_5" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="7426" name="__module.model.gpt_neox.layers.30.attention/aten::size/Gather_5" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="4903" name="__module.model.gpt_neox.layers.30.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
+			<data approximation_mode="ERF" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="4832,4837" />
+				<port id="1" precision="FP32" names="3839">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>10240</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="7427" name="__module.model.gpt_neox.layers.30.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
+		<layer id="4904" name="self.model.gpt_neox.layers.30.mlp.dense_4h_to_h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 10240" offset="2519749200" size="26214400" />
 			<output>
-				<port id="2" precision="I64" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="7428" name="__module.model.gpt_neox.layers.30.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1">
+		<layer id="4905" name="Convert_74794" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64" names="4834,4835,4839,4840" />
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="7429" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Constant_1" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4906" name="self.model.gpt_neox.layers.30.mlp.dense_4h_to_h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="2545963600" size="2560" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="7430" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1">
+		<layer id="4907" name="Convert_74797" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
+				<port id="0" precision="U8">
+					<dim>2560</dim>
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7431" name="Constant_280312" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
+		<layer id="4908" name="self.model.gpt_neox.layers.30.mlp.dense_4h_to_h.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="7432" name="ScatterUpdate_280315" type="ScatterUpdate" version="opset3">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
+		<layer id="4909" name="self.model.gpt_neox.layers.30.mlp.dense_4h_to_h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="2545966160" size="5120" />
+			<output>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+			</output>
+		</layer>
+		<layer id="4910" name="self.model.gpt_neox.layers.30.mlp.dense_4h_to_h.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="1" precision="FP16">
+					<dim>2560</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7433" name="Constant_280318" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
+		<layer id="4911" name="__module.model.gpt_neox.layers.30.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
+			<data destination_type="f32" />
 			<rt_info>
-				<attribute name="precise" version="0" />
+				<attribute name="decompression" version="0" />
 			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+			<input>
+				<port id="0" precision="FP16">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="7434" name="Constant_280321" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7435" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4912" name="__module.model.gpt_neox.layers.30.mlp.dense_4h_to_h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>10240</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>2560</dim>
+					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4841,x2.121">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7436" name="Constant_299376_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
+		<layer id="4913" name="self.model.gpt_neox.layers.30.mlp.dense_4h_to_h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="2560" offset="2545971280" size="5120" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.30.mlp.dense_4h_to_h.bias">
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7437" name="Constant_299376" type="Convert" version="opset1">
+		<layer id="4914" name="__module.model.gpt_neox.layers.30.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
 			</rt_info>
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16">
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7438" name="__module.model.gpt_neox.layers.30.attention/aten::neg/Multiply" type="Multiply" version="opset1">
+		<layer id="4915" name="__module.model.gpt_neox.layers.30.mlp.dense_4h_to_h/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4842">
+				<port id="2" precision="FP32" names="3842,input.123">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7439" name="Constant_280366" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7440" name="Constant_280371" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4916" name="Constant_37530" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2545976400" size="10240" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7441" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4917" name="__module.model.gpt_neox.layers.30.input_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4836">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7442" name="__module.model.gpt_neox.layers.30.attention/aten::cat/Concat" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="4918" name="Constant_37531" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2545986640" size="10240" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4919" name="__module.model.gpt_neox.layers.30.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4844">
+				<port id="2" precision="FP32" names="3745,hidden_states.121">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7443" name="self.model.gpt_neox.layers.30.attention.rotary_emb.sin_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201162356" size="81920" />
+		<layer id="4920" name="self.model.gpt_neox.layers.30.attention.query_key_value.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 2560" offset="2545996880" size="19660800" />
 			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.30.attention.rotary_emb.sin_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7444" name="self.model.gpt_neox.layers.30.attention.rotary_emb.sin_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="4921" name="Convert_75154" type="Convert" version="opset1">
+			<data destination_type="f16" />
 			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7445" name="__module.model.gpt_neox.layers.30.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4922" name="self.model.gpt_neox.layers.30.attention.query_key_value.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 1" offset="2565657680" size="7680" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="U8">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7446" name="__module.model.gpt_neox.layers.30.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4923" name="Convert_75157" type="Convert" version="opset1">
+			<data destination_type="f16" />
+			<input>
+				<port id="0" precision="U8">
+					<dim>7680</dim>
+					<dim>1</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7447" name="__module.model.gpt_neox.layers.30.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4924" name="self.model.gpt_neox.layers.30.attention.query_key_value.weight/zero_point/subtract" type="Subtract" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
+			</input>
+			<output>
+				<port id="2" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4925" name="self.model.gpt_neox.layers.30.attention.query_key_value.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="7680, 1" offset="2565665360" size="15360" />
+			<output>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I64">
+			</output>
+		</layer>
+		<layer id="4926" name="self.model.gpt_neox.layers.30.attention.query_key_value.weight/fq_weights_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP16">
+					<dim>7680</dim>
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4821,4824,sin.121">
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="2" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7448" name="Constant_38380" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
+		<layer id="4927" name="self.model.gpt_neox.layers.30.attention.query_key_value.weight/fq_weights_1/convert" type="Convert" version="opset1">
+			<data destination_type="f32" />
+			<input>
+				<port id="0" precision="FP16">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I32" />
+				<port id="1" precision="FP32">
+					<dim>7680</dim>
+					<dim>2560</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="7449" name="__module.model.gpt_neox.layers.30.attention/aten::index/Gather_1" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="4928" name="__module.model.gpt_neox.layers.30.attention.query_key_value/ov_ext::linear/MatMul" type="MatMul" version="opset1">
+			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>2560</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>7680</dim>
+					<dim>2560</dim>
 				</port>
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="4829">
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7450" name="__module.model.gpt_neox.layers.30.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I64" />
-			</input>
+		<layer id="4929" name="Constant_37532" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 7680" offset="2565680720" size="30720" />
 			<output>
-				<port id="2" precision="FP32" names="4830,sin.123">
-					<dim>-1</dim>
+				<port id="0" precision="FP32">
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>1</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7451" name="__module.model.gpt_neox.layers.30.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
+		<layer id="4930" name="__module.model.gpt_neox.layers.30.attention.query_key_value/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>7680</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>1</dim>
+					<dim>7680</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4845">
+				<port id="2" precision="FP32" names="3752,qkv.121">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7452" name="__module.model.gpt_neox.layers.30.attention/aten::add/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="4931" name="__module.model.gpt_neox.layers.30.attention/prim::ListConstruct" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201115992" size="32" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4932" name="__module.model.gpt_neox.layers.30.attention/aten::view/Reshape" type="Reshape" version="opset1">
+			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>7680</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4846">
+				<port id="2" precision="FP32" names="3756,qkv.123">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7453" name="Constant_280417" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4933" name="Constant_19616" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7454" name="Constant_280420" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4934" name="Constant_19618" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7455" name="Constant_280423" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4935" name="Constant_19620" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7456" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4936" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Slice" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4808">
+				<port id="5" precision="FP32" names="3757">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>60</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7457" name="__module.model.gpt_neox.layers.30.attention/aten::cat/Concat_1" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="4937" name="Constant_19629" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
+			<output>
+				<port id="0" precision="I64" names="3758">
+					<dim>4</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4938" name="__module.model.gpt_neox.layers.30.attention/aten::permute/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>60</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4864,4874,query.183,query.185">
+				<port id="2" precision="FP32" names="3759,query.61">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -114142,41 +75107,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7458" name="Constant_280429" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4939" name="Constant_19670" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7459" name="Constant_280432" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4940" name="Constant_19672" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7460" name="Constant_280435" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4941" name="Constant_19674" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7461" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4942" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Slice_1" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -114185,17 +75140,20 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4809,k.61">
+				<port id="5" precision="FP32" names="3766,q.61">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -114203,7 +75161,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7462" name="__module.model.gpt_neox.layers.30.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
+		<layer id="4943" name="__module.model.gpt_neox.layers.30.attention/aten::mul/Multiply" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -114220,7 +75178,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4847">
+				<port id="2" precision="FP32" names="3772">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -114228,160 +75186,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7463" name="Constant_280439" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7464" name="Constant_280438" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4944" name="Constant_19828" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7465" name="__module.model.gpt_neox.layers.30.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7466" name="__module.model.gpt_neox.layers.30.attention/aten::size/Constant_7" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4945" name="Constant_19830" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="7467" name="__module.model.gpt_neox.layers.30.attention/aten::size/Gather_7" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="4848,4853" />
-			</output>
-		</layer>
-		<layer id="7468" name="__module.model.gpt_neox.layers.30.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" />
-			</output>
-		</layer>
-		<layer id="7469" name="__module.model.gpt_neox.layers.30.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-			</input>
-			<output>
-				<port id="1" precision="I64" names="4850,4851,4855,4856" />
-			</output>
-		</layer>
-		<layer id="7470" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="7471" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Unsqueeze_33" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7472" name="Constant_280437" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4946" name="Constant_19832" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7473" name="ScatterUpdate_280440" type="ScatterUpdate" version="opset3">
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7474" name="Constant_280443" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7475" name="Constant_280446" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
 			</output>
 		</layer>
-		<layer id="7476" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4947" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Slice_2" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -114390,17 +75219,20 @@
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4857,x2.123">
+				<port id="5" precision="FP32" names="3782,x2.121">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -114408,32 +75240,10 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7477" name="Constant_299377_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7478" name="Constant_299377" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
+		<layer id="4948" name="Constant_37533" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>1</dim>
@@ -114441,7 +75251,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7479" name="__module.model.gpt_neox.layers.30.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
+		<layer id="4949" name="__module.model.gpt_neox.layers.30.attention/aten::neg/Multiply" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -114458,7 +75268,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4858">
+				<port id="2" precision="FP32" names="3783">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -114466,30 +75276,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7480" name="Constant_280491" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4950" name="Constant_19725" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7481" name="Constant_280496" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4951" name="Constant_19769" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4952" name="Constant_19771" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7482" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4953" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Slice_3" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -114498,17 +75309,20 @@
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4852">
+				<port id="5" precision="FP32" names="3777">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -114516,7 +75330,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7483" name="__module.model.gpt_neox.layers.30.attention/aten::cat/Concat_2" type="Concat" version="opset1">
+		<layer id="4954" name="__module.model.gpt_neox.layers.30.attention/aten::cat/Concat" type="Concat" version="opset1">
 			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
@@ -114533,7 +75347,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4860">
+				<port id="2" precision="FP32" names="3785">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -114541,7 +75355,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7484" name="__module.model.gpt_neox.layers.30.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
+		<layer id="4955" name="__module.model.gpt_neox.layers.30.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -114558,7 +75372,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4861">
+				<port id="2" precision="FP32" names="3786">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -114566,7 +75380,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7485" name="__module.model.gpt_neox.layers.30.attention/aten::add/Add_1" type="Add" version="opset1">
+		<layer id="4956" name="__module.model.gpt_neox.layers.30.attention/aten::add/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -114583,7 +75397,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4862">
+				<port id="2" precision="FP32" names="3787">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -114591,41 +75405,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7486" name="Constant_280541" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4957" name="Constant_19680" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7487" name="Constant_280544" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4958" name="Constant_19682" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7488" name="Constant_280547" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4959" name="Constant_19684" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7489" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4960" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Slice_4" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -114634,17 +75438,20 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4810">
+				<port id="5" precision="FP32" names="3767">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -114652,7 +75459,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7490" name="__module.model.gpt_neox.layers.30.attention/aten::cat/Concat_3" type="Concat" version="opset1">
+		<layer id="4961" name="__module.model.gpt_neox.layers.30.attention/aten::cat/Concat_1" type="Concat" version="opset1">
 			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
@@ -114669,32 +75476,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4866">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7491" name="__module.model.gpt_neox.layers.30.attention/aten::to/Convert_1" type="Concat" version="opset1">
-			<data axis="-2" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="present.30.key">
+				<port id="2" precision="FP32" names="3805">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -114702,49 +75484,37 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7492" name="Constant_338222_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
+		<layer id="4962" name="Constant_53271" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="7493" name="Constant_338222" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
+				<port id="0" precision="FP32" />
 			</output>
 		</layer>
-		<layer id="7494" name="Constant_338215" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
+		<layer id="4963" name="Constant_53264" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7495" name="Constant_338217" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
+		<layer id="4964" name="Constant_53266" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7496" name="Constant_338219" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
+		<layer id="4965" name="Constant_53268" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7497" name="Concat_338220" type="Concat" version="opset1">
+		<layer id="4966" name="Concat_53269" type="Concat" version="opset1">
 			<data axis="0" />
 			<input>
 				<port id="0" precision="I64">
@@ -114766,7 +75536,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7498" name="Broadcast_338223" type="Broadcast" version="opset3">
+		<layer id="4967" name="Broadcast_53272" type="Broadcast" version="opset3">
 			<data mode="numpy" />
 			<input>
 				<port id="0" precision="FP32" />
@@ -114783,8 +75553,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7499" name="ReadValue_330523" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.30.valuepresent.30.value" variable_type="f32" variable_shape="?,32,?,80" />
+		<layer id="4968" name="ReadValue_50279" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.30.keypresent.30.key" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -114794,7 +75564,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="past_key_values.30.value">
+				<port id="1" precision="FP32" names="past_key_values.30.key">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -114802,13 +75572,13 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7500" name="Constant_323568" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
+		<layer id="4969" name="Constant_47958" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="7501" name="Gather_323569" type="Gather" version="opset8">
+		<layer id="4970" name="Gather_47959" type="Gather" version="opset8">
 			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="FP32">
@@ -114831,41 +75601,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7502" name="Constant_280553" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4971" name="Constant_19634" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7503" name="Constant_280556" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4972" name="Constant_19636" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7504" name="Constant_280559" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4973" name="Constant_19638" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7505" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4974" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Slice_5" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -114874,17 +75634,20 @@
 					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4804">
+				<port id="5" precision="FP32" names="3760">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -114892,15 +75655,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7506" name="Constant_17681" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="4975" name="Constant_19647" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64" names="4805">
+				<port id="0" precision="I64" names="3761">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7507" name="__module.model.gpt_neox.layers.30.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
+		<layer id="4976" name="__module.model.gpt_neox.layers.30.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -114913,7 +75676,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4806">
+				<port id="2" precision="FP32" names="3762,key.121">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -114921,8 +75684,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7508" name="__module.model.gpt_neox.layers.30.attention/aten::cat/Concat_5" type="Concat" version="opset1">
-			<data axis="-2" />
+		<layer id="4977" name="Constant_19690" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4978" name="Constant_19692" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4979" name="Constant_19694" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4980" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Slice_6" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -114930,176 +75716,174 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
+			</input>
+			<output>
+				<port id="5" precision="FP32" names="3768,k.61">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="4981" name="__module.model.gpt_neox.layers.30.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="present.30.value">
+				<port id="2" precision="FP32" names="3788">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7509" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Slice_11" type="Const" version="opset1">
-			<data element_type="boolean" shape="1, 1, 2048, 2048" offset="201244276" size="4194304" />
+		<layer id="4982" name="Constant_19971" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
-				<port id="0" precision="BOOL" names="4883">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7510" name="Constant_280563" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4983" name="Constant_19973" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>3</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7511" name="Constant_280562" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162298" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4984" name="Constant_19975" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7512" name="__module.model.gpt_neox.layers.30.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="4985" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Slice_7" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="5" precision="FP32" names="3798,x2.123">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7513" name="__module.model.gpt_neox.layers.30.attention/aten::size/Constant_8" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4986" name="Constant_37534" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="7514" name="__module.model.gpt_neox.layers.30.attention/aten::size/Gather_8" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="4878" />
 			</output>
 		</layer>
-		<layer id="7515" name="__module.model.gpt_neox.layers.30.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="4987" name="__module.model.gpt_neox.layers.30.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>10</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="3799">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7516" name="__module.model.gpt_neox.layers.30.attention/aten::size/Constant_9" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4988" name="Constant_19868" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="7517" name="__module.model.gpt_neox.layers.30.attention/aten::size/Gather_9" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="4876" />
-			</output>
-		</layer>
-		<layer id="7518" name="__module.model.gpt_neox.layers.30.attention/aten::sub/Subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" names="4880,4881" />
-			</output>
-		</layer>
-		<layer id="7519" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Constant_4" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
 			</output>
 		</layer>
-		<layer id="7520" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="4989" name="Constant_19912" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7521" name="Constant_280561" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4990" name="Constant_19914" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7522" name="ScatterUpdate_280564" type="ScatterUpdate" version="opset3">
+		<layer id="4991" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Slice_8" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -115107,146 +75891,236 @@
 				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
+				<port id="5" precision="FP32" names="3793">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7523" name="Constant_280565" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4992" name="__module.model.gpt_neox.layers.30.attention/aten::cat/Concat_2" type="Concat" version="opset1">
+			<data axis="-1" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="2" precision="FP32" names="3801">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7524" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1">
+		<layer id="4993" name="__module.model.gpt_neox.layers.30.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
 			</input>
 			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="3802">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7525" name="ScatterUpdate_280566" type="ScatterUpdate" version="opset3">
+		<layer id="4994" name="__module.model.gpt_neox.layers.30.attention/aten::add/Add_1" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
-				<port id="0" precision="I64">
-					<dim>3</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="3803">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
-				<port id="3" precision="I32">
+			</output>
+		</layer>
+		<layer id="4995" name="Constant_19700" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
+			</output>
+		</layer>
+		<layer id="4996" name="Constant_19702" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7526" name="Constant_280569" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438612" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4997" name="Constant_19704" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>3</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7527" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="4998" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Slice_9" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>3</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>3</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>3</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="BOOL" names="4884">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="5" precision="FP32" names="3769">
 					<dim>-1</dim>
-					<dim>2048</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7528" name="Constant_280642" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="4999" name="__module.model.gpt_neox.layers.30.attention/aten::cat/Concat_3" type="Concat" version="opset1">
+			<data axis="-1" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>60</dim>
+				</port>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="3807">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7529" name="Constant_280643" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="5000" name="__module.model.gpt_neox.layers.30.attention/aten::cat/Concat_4" type="Concat" version="opset1">
+			<data axis="-2" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32" names="present.30.key">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="5001" name="Constant_53283" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
+			<output>
+				<port id="0" precision="FP32" />
+			</output>
+		</layer>
+		<layer id="5002" name="Constant_53276" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7530" name="Constant_280639" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="5003" name="Constant_53278" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7531" name="Constant_280638" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="5004" name="Constant_53280" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7532" name="ScatterUpdate_280644" type="ScatterUpdate" version="opset3">
+		<layer id="5005" name="Concat_53281" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -115254,7 +76128,7 @@
 				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
@@ -115264,123 +76138,195 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7533" name="Constant_280647" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
+		<layer id="5006" name="Broadcast_53284" type="Broadcast" version="opset3">
+			<data mode="numpy" />
+			<input>
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
+			</input>
+			<output>
+				<port id="2" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="7534" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="5007" name="ReadValue_50281" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.30.valuepresent.30.value" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
-				<port id="0" precision="BOOL">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
+				</port>
+			</input>
+			<output>
+				<port id="1" precision="FP32" names="past_key_values.30.value">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="5008" name="Constant_47961" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64" />
+			</output>
+		</layer>
+		<layer id="5009" name="Gather_47962" type="Gather" version="opset8">
+			<data batch_dims="0" />
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I32">
+					<dim>-1</dim>
+				</port>
+				<port id="2" precision="I64" />
+			</input>
+			<output>
+				<port id="3" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="5010" name="Constant_19652" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="5011" name="Constant_19654" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="5012" name="Constant_19656" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="5013" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Slice_10" type="Slice" version="opset8">
+			<input>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>2048</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="BOOL" names="4885,4886">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="5" precision="FP32" names="3763">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7535" name="__module.model.gpt_neox.layers.30.attention/aten::where/Convert_1_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="7536" name="__module.model.gpt_neox.layers.30.attention/aten::where/Convert_1" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="7537" name="__module.model.gpt_neox.layers.30.attention/aten::full/Convert" type="Const" version="opset1">
-			<data element_type="f32" shape="" offset="205438636" size="4" />
+		<layer id="5014" name="Constant_19665" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="FP32" names="4873" />
+				<port id="0" precision="I64" names="3764">
+					<dim>4</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="7538" name="__module.model.gpt_neox.layers.30.attention/aten::where/Select" type="Select" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="5015" name="__module.model.gpt_neox.layers.30.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
 			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
-				<port id="1" precision="FP32" />
-				<port id="2" precision="FP32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="4887,causal_mask.121">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="3765">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7539" name="Add_290518" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="5016" name="__module.model.gpt_neox.layers.30.attention/aten::cat/Concat_5" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="FP32">
 					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
+				<port id="2" precision="FP32" names="present.30.value">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="5017" name="Constant_20047" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7540" name="ShapeOf_290519" type="ShapeOf" version="opset3">
+		<layer id="5018" name="ShapeOf_40842" type="ShapeOf" version="opset3">
 			<data output_type="i64" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</input>
 			<output>
@@ -115389,27 +76335,21 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7541" name="Constant_254779" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="5019" name="Constant_40843" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116168" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7542" name="__module.model.gpt_neox.layers.30.attention/aten::size/Constant_10" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="5020" name="Constant_40844" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="I64" />
 			</output>
 		</layer>
-		<layer id="7543" name="__module.model.gpt_neox.layers.30.attention/aten::size/Gather_10" type="Gather" version="opset8">
+		<layer id="5021" name="Gather_40845" type="Gather" version="opset8">
 			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="I64">
@@ -115418,16 +76358,16 @@
 				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I32" />
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="4871">
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7544" name="__module.model.gpt_neox.layers.30.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1">
-			<data axis="0" />
+		<layer id="5022" name="Add_40848" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="I64">
 					<dim>1</dim>
@@ -115435,49 +76375,22 @@
 				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
 			</input>
 			<output>
-				<port id="4" precision="I64" names="4888">
-					<dim>4</dim>
+				<port id="2" precision="I64" names="3812">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7545" name="__module.model.gpt_neox.layers.30.attention/aten::expand/Abs" type="Abs" version="opset1">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
+		<layer id="5023" name="Constant_20050" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7546" name="Maximum_290520" type="Maximum" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7547" name="__module.model.gpt_neox.layers.30.attention/aten::add/Add_2" type="Broadcast" version="opset3">
-			<data mode="numpy" />
+		<layer id="5024" name="__module.model.gpt_neox.layers.30.attention/aten::slice/Slice_14" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -115486,19 +76399,28 @@
 					<dim>-1</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4890">
-					<dim>-1</dim>
+				<port id="5" precision="FP32" names="3816">
 					<dim>-1</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
 					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7548" name="__module.model.gpt_neox.layers.30.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+		<layer id="5025" name="__module.model.gpt_neox.layers.30.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
 			<data causal="false" />
 			<input>
 				<port id="0" precision="FP32">
@@ -115521,13 +76443,13 @@
 				</port>
 				<port id="3" precision="FP32">
 					<dim>-1</dim>
-					<dim>-1</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
 					<dim>-1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4891,4892,sdpa_result.61,tensor.121">
+				<port id="4" precision="FP32" names="3817,attn_output.181">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -115535,15 +76457,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7549" name="Constant_18122" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="5026" name="__module.model.gpt_neox.layers.30.attention/aten::transpose/Constant" type="Const" version="opset1">
+			<data element_type="i32" shape="4" offset="201116200" size="16" />
 			<output>
-				<port id="0" precision="I64" names="4893">
+				<port id="0" precision="I32">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7550" name="__module.model.gpt_neox.layers.30.attention/aten::permute/Transpose_3" type="Transpose" version="opset1">
+		<layer id="5027" name="__module.model.gpt_neox.layers.30.attention/aten::transpose/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -115551,12 +76473,12 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="I32">
 					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4894">
+				<port id="2" precision="FP32" names="3818">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -115564,18 +76486,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7551" name="__module.model.gpt_neox.layers.30.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438650" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="5028" name="Constant_41014" type="Const" version="opset1">
+			<data element_type="i64" shape="3" offset="201116216" size="24" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>3</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7552" name="__module.model.gpt_neox.layers.30.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
+		<layer id="5029" name="__module.model.gpt_neox.layers.30.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
 			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -115589,15 +76508,15 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4899">
+				<port id="2" precision="FP32" names="3821">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7553" name="self.model.gpt_neox.layers.30.attention.dense.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 2560" offset="2568805074" size="6553600" />
+		<layer id="5030" name="self.model.gpt_neox.layers.30.attention.dense.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 2560" offset="2565711440" size="6553600" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>2560</dim>
@@ -115605,7 +76524,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7554" name="Convert_22634" type="Convert" version="opset1">
+		<layer id="5031" name="Convert_75506" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -115620,8 +76539,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7555" name="self.model.gpt_neox.layers.30.attention.dense.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="2575358674" size="2560" />
+		<layer id="5032" name="self.model.gpt_neox.layers.30.attention.dense.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="2572265040" size="2560" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>2560</dim>
@@ -115629,7 +76548,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7556" name="Convert_22637" type="Convert" version="opset1">
+		<layer id="5033" name="Convert_75509" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -115644,7 +76563,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7557" name="self.model.gpt_neox.layers.30.attention.dense.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="5034" name="self.model.gpt_neox.layers.30.attention.dense.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -115663,8 +76582,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7558" name="self.model.gpt_neox.layers.30.attention.dense.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="2575361234" size="5120" />
+		<layer id="5035" name="self.model.gpt_neox.layers.30.attention.dense.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="2572267600" size="5120" />
 			<output>
 				<port id="0" precision="FP16">
 					<dim>2560</dim>
@@ -115672,7 +76591,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7559" name="self.model.gpt_neox.layers.30.attention.dense.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="5036" name="self.model.gpt_neox.layers.30.attention.dense.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -115691,11 +76610,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7560" name="self.model.gpt_neox.layers.30.attention.dense.weight" type="Convert" version="opset1">
+		<layer id="5037" name="self.model.gpt_neox.layers.30.attention.dense.weight/fq_weights_1/convert" type="Convert" version="opset1">
 			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
 			<input>
 				<port id="0" precision="FP16">
 					<dim>2560</dim>
@@ -115709,7 +76625,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7561" name="__module.model.gpt_neox.layers.30.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="5038" name="__module.model.gpt_neox.layers.30.attention.dense/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -115730,37 +76646,17 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7562" name="Constant_299378_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2575366354" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7563" name="Constant_299378" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="5039" name="Constant_37535" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2572272720" size="10240" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7564" name="__module.model.gpt_neox.layers.30.attention.dense/aten::linear/Add" type="Add" version="opset1">
+		<layer id="5040" name="__module.model.gpt_neox.layers.30.attention.dense/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -115775,14 +76671,14 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4902,4904,input.123">
+				<port id="2" precision="FP32" names="3824,3826,input.121">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7565" name="__module.model.gpt_neox.layers.30/aten::add/Add" type="Add" version="opset1">
+		<layer id="5041" name="__module.model.gpt_neox.layers.30/aten::add/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -115797,14 +76693,14 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4922">
+				<port id="2" precision="FP32" names="3844">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7566" name="__module.model.gpt_neox.layers.30/aten::add/Add_1" type="Add" version="opset1">
+		<layer id="5042" name="__module.model.gpt_neox.layers.30/aten::add/Add_1" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -115819,22 +76715,22 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4923,4925,hidden_states">
+				<port id="2" precision="FP32" names="3845,3847,hidden_states">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7567" name="__module.model.gpt_neox.layers.31.input_layernorm/aten::layer_norm/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867670" size="4" />
+		<layer id="5043" name="__module.model.gpt_neox.layers.31.input_layernorm/aten::layer_norm/Multiply" type="Const" version="opset1">
+			<data element_type="i32" shape="1" offset="128867668" size="4" />
 			<output>
 				<port id="0" precision="I32">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7568" name="__module.model.gpt_neox.layers.31.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
+		<layer id="5044" name="__module.model.gpt_neox.layers.31.input_layernorm/aten::layer_norm/MVN" type="MVN" version="opset6">
 			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
 			<input>
 				<port id="0" precision="FP32">
@@ -115854,37 +76750,17 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7569" name="Constant_299379_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2575371474" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7570" name="Constant_299379" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="5045" name="Constant_37536" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2572282960" size="10240" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7571" name="__module.model.gpt_neox.layers.31.input_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
+		<layer id="5046" name="__module.model.gpt_neox.layers.31.input_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -115906,37 +76782,17 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7572" name="Constant_299380_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2575376594" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7573" name="Constant_299380" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="5047" name="Constant_37537" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2572293200" size="10240" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7574" name="__module.model.gpt_neox.layers.31.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+		<layer id="5048" name="__module.model.gpt_neox.layers.31.input_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -115951,15 +76807,15 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4935">
+				<port id="2" precision="FP32" names="3857,hidden_states.125">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7575" name="self.model.gpt_neox.layers.31.attention.query_key_value.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 2560" offset="2575381714" size="19660800" />
+		<layer id="5049" name="self.model.gpt_neox.layers.31.attention.query_key_value.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 2560" offset="2572303440" size="19660800" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>7680</dim>
@@ -115967,7 +76823,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7576" name="Convert_22664" type="Convert" version="opset1">
+		<layer id="5050" name="Convert_75165" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -115982,8 +76838,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7577" name="self.model.gpt_neox.layers.31.attention.query_key_value.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="7680, 1" offset="2595042514" size="7680" />
+		<layer id="5051" name="self.model.gpt_neox.layers.31.attention.query_key_value.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="7680, 1" offset="2591964240" size="7680" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>7680</dim>
@@ -115991,7 +76847,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7578" name="Convert_22667" type="Convert" version="opset1">
+		<layer id="5052" name="Convert_75168" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -116006,7 +76862,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7579" name="self.model.gpt_neox.layers.31.attention.query_key_value.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="5053" name="self.model.gpt_neox.layers.31.attention.query_key_value.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -116025,8 +76881,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7580" name="self.model.gpt_neox.layers.31.attention.query_key_value.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="7680, 1" offset="2595050194" size="15360" />
+		<layer id="5054" name="self.model.gpt_neox.layers.31.attention.query_key_value.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="7680, 1" offset="2591971920" size="15360" />
 			<output>
 				<port id="0" precision="FP16">
 					<dim>7680</dim>
@@ -116034,7 +76890,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7581" name="self.model.gpt_neox.layers.31.attention.query_key_value.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="5055" name="self.model.gpt_neox.layers.31.attention.query_key_value.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -116053,11 +76909,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7582" name="self.model.gpt_neox.layers.31.attention.query_key_value.weight" type="Convert" version="opset1">
+		<layer id="5056" name="self.model.gpt_neox.layers.31.attention.query_key_value.weight/fq_weights_1/convert" type="Convert" version="opset1">
 			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
 			<input>
 				<port id="0" precision="FP16">
 					<dim>7680</dim>
@@ -116071,7 +76924,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7583" name="__module.model.gpt_neox.layers.31.attention.query_key_value/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="5057" name="__module.model.gpt_neox.layers.31.attention.query_key_value/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -116092,563 +76945,152 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7584" name="Constant_299381_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 7680" offset="2595065554" size="15360" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7585" name="Constant_299381" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</input>
+		<layer id="5058" name="Constant_37538" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 7680" offset="2591987280" size="30720" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7586" name="__module.model.gpt_neox.layers.31.attention.query_key_value/aten::linear/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
 				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-				<port id="1" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>7680</dim>
 				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="4942,qkv.125">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>7680</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7587" name="__module.model.gpt_neox.layers.31.attention/prim::ListConstruct/Concat" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080154" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
 			</output>
 		</layer>
-		<layer id="7588" name="__module.model.gpt_neox.layers.31.attention/aten::view/Reshape" type="Reshape" version="opset1">
-			<data special_zero="true" />
+		<layer id="5059" name="__module.model.gpt_neox.layers.31.attention.query_key_value/ov_ext::linear/Add" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>7680</dim>
 				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="4946,qkv">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7589" name="Constant_280654" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7590" name="Constant_280657" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7591" name="Constant_280660" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7592" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="FP32" names="4953">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7593" name="Constant_18262" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
-			<output>
-				<port id="0" precision="I64" names="4954">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7594" name="__module.model.gpt_neox.layers.31.attention/aten::permute/Transpose" type="Transpose" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="4955">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7595" name="__module.model.gpt_neox.layers.31.attention/aten::cat/Concat" type="Concat" version="opset1">
-			<data axis="-2" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32" names="present.31.value">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7597" name="Constant_338235_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="7598" name="Constant_338235" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
-			</input>
-			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="7599" name="Constant_338228" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="10" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7600" name="Constant_338230" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7601" name="Constant_338232" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="18" size="8" />
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7602" name="Concat_338233" type="Concat" version="opset1">
-			<data axis="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7603" name="Broadcast_338236" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32" />
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7604" name="ReadValue_330525" type="ReadValue" version="opset6">
-			<data variable_id="past_key_values.31.keypresent.31.key" variable_type="f32" variable_shape="?,32,?,80" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>0</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32" names="past_key_values.31.key">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7605" name="Constant_323571" type="Const" version="opset1">
-			<data element_type="i64" shape="" offset="2" size="8" />
-			<output>
-				<port id="0" precision="I64" />
-			</output>
-		</layer>
-		<layer id="7606" name="Gather_323572" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I64" />
-			</input>
-			<output>
-				<port id="3" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7607" name="Constant_280666" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7608" name="Constant_280669" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162266" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7609" name="Constant_280672" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7610" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>240</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>7680</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4950">
+				<port id="2" precision="FP32" names="3864,qkv.125">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
+					<dim>7680</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7611" name="Constant_18248" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="5060" name="__module.model.gpt_neox.layers.31.attention/prim::ListConstruct" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201115992" size="32" />
 			<output>
-				<port id="0" precision="I64" names="4951">
+				<port id="0" precision="I64">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7612" name="__module.model.gpt_neox.layers.31.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
+		<layer id="5061" name="__module.model.gpt_neox.layers.31.attention/aten::view/Reshape" type="Reshape" version="opset1">
+			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>80</dim>
+					<dim>7680</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4952,key.187">
+				<port id="2" precision="FP32" names="3868,qkv">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7613" name="Constant_280678" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="5062" name="Constant_20282" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7614" name="Constant_280681" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="5063" name="Constant_20284" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7615" name="Constant_280684" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="5064" name="Constant_20286" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7616" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Slice_2" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="5065" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Slice" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>80</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4958,k">
+				<port id="5" precision="FP32" names="3875">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7617" name="self.model.gpt_neox.layers.31.attention.rotary_emb.cos_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201080346" size="81920" />
-			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.31.attention.rotary_emb.cos_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7618" name="self.model.gpt_neox.layers.31.attention.rotary_emb.cos_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7619" name="__module.model.gpt_neox.layers.31.attention.rotary_emb/aten::slice/Unsqueeze" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="5066" name="Constant_20295" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
+				<port id="0" precision="I64" names="3876">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7620" name="__module.model.gpt_neox.layers.31.attention/aten::size/ShapeOf_2" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="5067" name="__module.model.gpt_neox.layers.31.attention/aten::permute/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>80</dim>
 				</port>
-			</input>
-			<output>
 				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="7621" name="__module.model.gpt_neox.layers.31.attention/aten::size/Constant_2" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="7622" name="__module.model.gpt_neox.layers.31.attention/aten::size/Gather_2" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="I64" names="4960" />
+				<port id="2" precision="FP32" names="3877">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
 			</output>
 		</layer>
-		<layer id="7623" name="__module.model.gpt_neox.layers.31.attention/aten::size/ShapeOf_3" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="5068" name="__module.model.gpt_neox.layers.31.attention/aten::cat/Concat" type="Concat" version="opset1">
+			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -116656,82 +77098,57 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
+				<port id="1" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
 			</input>
 			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
+				<port id="2" precision="FP32" names="present.31.value">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7624" name="__module.model.gpt_neox.layers.31.attention/aten::size/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="5070" name="Constant_53295" type="Const" version="opset1">
+			<data element_type="f32" shape="" offset="0" size="4" />
 			<output>
-				<port id="0" precision="I32" />
+				<port id="0" precision="FP32" />
 			</output>
 		</layer>
-		<layer id="7625" name="__module.model.gpt_neox.layers.31.attention/aten::size/Gather_3" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
+		<layer id="5071" name="Constant_53288" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="12" size="8" />
+			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="4962" />
-			</output>
-		</layer>
-		<layer id="7626" name="__module.model.gpt_neox.layers.31.attention/aten::add_/Add" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" names="4961,4965,4966,seq_len.125" />
-			</output>
-		</layer>
-		<layer id="7627" name="__module.model.gpt_neox.layers.31.attention.rotary_emb/aten::slice/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
 			</output>
 		</layer>
-		<layer id="7628" name="__module.model.gpt_neox.layers.31.attention.rotary_emb/aten::slice/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
+		<layer id="5072" name="Constant_53290" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="2" precision="I64">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7629" name="__module.model.gpt_neox.layers.31.attention.rotary_emb/aten::slice/Unsqueeze_2" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="5073" name="Constant_53292" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7630" name="__module.model.gpt_neox.layers.31.attention.rotary_emb/aten::slice/Slice" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="5074" name="Concat_53293" type="Concat" version="opset1">
+			<data axis="0" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -116744,195 +77161,107 @@
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4969,4972,cos.125">
-					<dim>-1</dim>
-					<dim>20</dim>
+				<port id="4" precision="I64">
+					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7631" name="Constant_38383" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="7632" name="__module.model.gpt_neox.layers.31.attention/aten::index/Gather" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="5075" name="Broadcast_53296" type="Broadcast" version="opset3">
+			<data mode="numpy" />
 			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
+				<port id="0" precision="FP32" />
+				<port id="1" precision="I64">
+					<dim>4</dim>
 				</port>
-				<port id="2" precision="I32" />
 			</input>
 			<output>
-				<port id="3" precision="FP32" names="4975">
-					<dim>-1</dim>
+				<port id="2" precision="FP32">
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7633" name="__module.model.gpt_neox.layers.31.attention/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1">
+		<layer id="5076" name="ReadValue_50283" type="ReadValue" version="opset6">
+			<data variable_id="past_key_values.31.keypresent.31.key" variable_type="f32" variable_shape="?,32,?,80" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>0</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4976,cos">
+				<port id="1" precision="FP32" names="past_key_values.31.key">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7634" name="__module.model.gpt_neox.layers.31.attention/aten::mul/Multiply" type="Multiply" version="opset1">
-			<data auto_broadcast="numpy" />
+		<layer id="5077" name="Constant_47964" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
+			<output>
+				<port id="0" precision="I64" />
+			</output>
+		</layer>
+		<layer id="5078" name="Gather_47965" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>80</dim>
 				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
+				<port id="1" precision="I32">
 					<dim>-1</dim>
-					<dim>20</dim>
 				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4996">
+				<port id="3" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7635" name="Constant_280689" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7636" name="Constant_280688" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="5079" name="Constant_20264" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7637" name="__module.model.gpt_neox.layers.31.attention/aten::size/ShapeOf_5" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7638" name="__module.model.gpt_neox.layers.31.attention/aten::size/Constant_5" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="5080" name="Constant_20266" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116152" size="8" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="7639" name="__module.model.gpt_neox.layers.31.attention/aten::size/Gather_5" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="4997,5002" />
-			</output>
-		</layer>
-		<layer id="7640" name="__module.model.gpt_neox.layers.31.attention/aten::floor_divide/Divide_1" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" />
-			</output>
-		</layer>
-		<layer id="7641" name="__module.model.gpt_neox.layers.31.attention/aten::floor_divide/Floor_1" type="Floor" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-			</input>
-			<output>
-				<port id="1" precision="I64" names="4999,5000,5004,5005" />
-			</output>
-		</layer>
-		<layer id="7642" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Constant_1" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="7643" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Unsqueeze_17" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7644" name="Constant_280687" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="5081" name="Constant_20268" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="I32">
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7645" name="ScatterUpdate_280690" type="ScatterUpdate" version="opset3">
+		<layer id="5082" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Slice_1" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
@@ -116940,148 +77269,155 @@
 				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="3" precision="I32">
+				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
 				<port id="4" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="7646" name="Constant_280693" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+			</input>
 			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
+				<port id="5" precision="FP32" names="3872">
+					<dim>-1</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7647" name="Constant_280696" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="5083" name="Constant_20277" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64">
+				<port id="0" precision="I64" names="3873">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7648" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Slice_3" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="5084" name="__module.model.gpt_neox.layers.31.attention/aten::permute/Transpose_1" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>4</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="5006,x2">
+				<port id="2" precision="FP32" names="3874,key.125">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>80</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7649" name="Constant_299382_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
+		<layer id="5085" name="Constant_20320" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="5086" name="Constant_20322" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="5087" name="Constant_20324" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7650" name="Constant_299382" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
+		<layer id="5088" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Slice_2" type="Slice" version="opset8">
 			<input>
-				<port id="0" precision="FP16">
+				<port id="0" precision="FP32">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>80</dim>
+				</port>
+				<port id="1" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="2" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="3" precision="I64">
 					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
 					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="5" precision="FP32" names="3880,k">
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7651" name="__module.model.gpt_neox.layers.31.attention/aten::neg/Multiply" type="Multiply" version="opset1">
+		<layer id="5089" name="__module.model.gpt_neox.layers.31.attention/aten::mul/Multiply" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>20</dim>
 				</port>
 				<port id="1" precision="FP32">
+					<dim>-1</dim>
 					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
+					<dim>-1</dim>
+					<dim>20</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="5007">
+				<port id="2" precision="FP32" names="3900">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7652" name="Constant_280741" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="5090" name="Constant_20601" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7653" name="Constant_280746" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="5091" name="Constant_20603" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="5092" name="Constant_20605" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7654" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Slice_4" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="5093" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Slice_3" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -117090,17 +77426,20 @@
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="5001">
+				<port id="5" precision="FP32" names="3910,x2">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -117108,8 +77447,19 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7655" name="__module.model.gpt_neox.layers.31.attention/aten::cat/Concat_1" type="Concat" version="opset1">
-			<data axis="-1" />
+		<layer id="5094" name="Constant_37539" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
+			<output>
+				<port id="0" precision="FP32">
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="5095" name="__module.model.gpt_neox.layers.31.attention/aten::neg/Multiply" type="Multiply" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -117118,75 +77468,51 @@
 					<dim>10</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>10</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="5009">
+				<port id="2" precision="FP32" names="3911">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7656" name="self.model.gpt_neox.layers.31.attention.rotary_emb.sin_cached_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="2048, 20" offset="201162356" size="81920" />
-			<output>
-				<port id="0" precision="FP16" names="self.model.gpt_neox.layers.31.attention.rotary_emb.sin_cached">
-					<dim>2048</dim>
-					<dim>20</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7657" name="self.model.gpt_neox.layers.31.attention.rotary_emb.sin_cached" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>2048</dim>
-					<dim>20</dim>
-				</port>
-			</input>
+		<layer id="5096" name="Constant_20498" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
-				<port id="1" precision="FP32">
-					<dim>2048</dim>
-					<dim>20</dim>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7658" name="__module.model.gpt_neox.layers.31.attention.rotary_emb/aten::slice/Unsqueeze_4" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="5097" name="Constant_20542" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7659" name="__module.model.gpt_neox.layers.31.attention.rotary_emb/aten::slice/Unsqueeze_6" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162306" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="5098" name="Constant_20544" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7660" name="__module.model.gpt_neox.layers.31.attention.rotary_emb/aten::slice/Slice_1" type="StridedSlice" version="opset1">
-			<data begin_mask="0" end_mask="0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="5099" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Slice_4" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
-					<dim>2048</dim>
+					<dim>-1</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
@@ -117198,60 +77524,45 @@
 				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
+				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4970,4973,sin.125">
+				<port id="5" precision="FP32" names="3905">
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>32</dim>
+					<dim>-1</dim>
+					<dim>10</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7661" name="Constant_38386" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="7662" name="__module.model.gpt_neox.layers.31.attention/aten::index/Gather_1" type="Gather" version="opset8">
-			<data batch_dims="0" />
+		<layer id="5100" name="__module.model.gpt_neox.layers.31.attention/aten::cat/Concat_1" type="Concat" version="opset1">
+			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-				<port id="1" precision="I32">
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="4978">
-					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>10</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="7663" name="__module.model.gpt_neox.layers.31.attention/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="FP32">
+				<port id="1" precision="FP32">
 					<dim>-1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
-					<dim>20</dim>
+					<dim>10</dim>
 				</port>
-				<port id="1" precision="I64" />
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4979,sin">
+				<port id="2" precision="FP32" names="3913">
 					<dim>-1</dim>
-					<dim>1</dim>
+					<dim>32</dim>
 					<dim>-1</dim>
 					<dim>20</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7664" name="__module.model.gpt_neox.layers.31.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
+		<layer id="5101" name="__module.model.gpt_neox.layers.31.attention/aten::mul/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -117268,7 +77579,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="5010">
+				<port id="2" precision="FP32" names="3914">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -117276,7 +77587,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7665" name="__module.model.gpt_neox.layers.31.attention/aten::add/Add" type="Add" version="opset1">
+		<layer id="5102" name="__module.model.gpt_neox.layers.31.attention/aten::add/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -117293,7 +77604,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="5011">
+				<port id="2" precision="FP32" names="3915">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -117301,41 +77612,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7666" name="Constant_280792" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="5103" name="Constant_20330" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7667" name="Constant_280795" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="5104" name="Constant_20332" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7668" name="Constant_280798" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="5105" name="Constant_20334" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7669" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Slice_5" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="5106" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Slice_5" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -117344,17 +77645,20 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4959">
+				<port id="5" precision="FP32" names="3881">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -117362,7 +77666,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7670" name="__module.model.gpt_neox.layers.31.attention/aten::cat/Concat_2" type="Concat" version="opset1">
+		<layer id="5107" name="__module.model.gpt_neox.layers.31.attention/aten::cat/Concat_2" type="Concat" version="opset1">
 			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
@@ -117379,7 +77683,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="5015">
+				<port id="2" precision="FP32" names="3919">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -117387,7 +77691,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7671" name="__module.model.gpt_neox.layers.31.attention/aten::to/Convert" type="Concat" version="opset1">
+		<layer id="5108" name="__module.model.gpt_neox.layers.31.attention/aten::cat/Concat_3" type="Concat" version="opset1">
 			<data axis="-2" />
 			<input>
 				<port id="0" precision="FP32">
@@ -117412,37 +77716,17 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7723" name="Constant_299383_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2595080914" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7724" name="Constant_299383" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="5160" name="Constant_37540" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2592018000" size="10240" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7725" name="__module.model.gpt_neox.layers.31.post_attention_layernorm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
+		<layer id="5161" name="__module.model.gpt_neox.layers.31.post_attention_layernorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -117464,37 +77748,17 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7726" name="Constant_299384_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2595086034" size="5120" />
+		<layer id="5162" name="Constant_37541" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2592028240" size="10240" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7727" name="Constant_299384" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7728" name="__module.model.gpt_neox.layers.31.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
+		<layer id="5163" name="__module.model.gpt_neox.layers.31.post_attention_layernorm/aten::layer_norm/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -117509,15 +77773,15 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="5060">
+				<port id="2" precision="FP32" names="3945">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7729" name="self.model.gpt_neox.layers.31.mlp.dense_h_to_4h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 2560" offset="2595091154" size="26214400" />
+		<layer id="5164" name="self.model.gpt_neox.layers.31.mlp.dense_h_to_4h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 2560" offset="2592038480" size="26214400" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>10240</dim>
@@ -117525,7 +77789,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7730" name="Convert_22684" type="Convert" version="opset1">
+		<layer id="5165" name="Convert_74804" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -117540,8 +77804,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7731" name="self.model.gpt_neox.layers.31.mlp.dense_h_to_4h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="10240, 1" offset="2621305554" size="10240" />
+		<layer id="5166" name="self.model.gpt_neox.layers.31.mlp.dense_h_to_4h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="10240, 1" offset="2618252880" size="10240" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>10240</dim>
@@ -117549,7 +77813,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7732" name="Convert_22687" type="Convert" version="opset1">
+		<layer id="5167" name="Convert_74807" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -117564,7 +77828,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7733" name="self.model.gpt_neox.layers.31.mlp.dense_h_to_4h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="5168" name="self.model.gpt_neox.layers.31.mlp.dense_h_to_4h.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -117583,8 +77847,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7734" name="self.model.gpt_neox.layers.31.mlp.dense_h_to_4h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="10240, 1" offset="2621315794" size="20480" />
+		<layer id="5169" name="self.model.gpt_neox.layers.31.mlp.dense_h_to_4h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="10240, 1" offset="2618263120" size="20480" />
 			<output>
 				<port id="0" precision="FP16">
 					<dim>10240</dim>
@@ -117592,7 +77856,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7735" name="self.model.gpt_neox.layers.31.mlp.dense_h_to_4h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="5170" name="self.model.gpt_neox.layers.31.mlp.dense_h_to_4h.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -117611,7 +77875,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7736" name="self.model.gpt_neox.layers.31.mlp.dense_h_to_4h.weight" type="Convert" version="opset1">
+		<layer id="5171" name="__module.model.gpt_neox.layers.31.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
@@ -117629,7 +77893,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7737" name="__module.model.gpt_neox.layers.31.mlp.dense_h_to_4h/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="5172" name="__module.model.gpt_neox.layers.31.mlp.dense_h_to_4h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -117650,37 +77914,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7738" name="Constant_299385_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 10240" offset="2621336274" size="20480" />
+		<layer id="5173" name="self.model.gpt_neox.layers.31.mlp.dense_h_to_4h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="10240" offset="2618283600" size="20480" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.31.mlp.dense_h_to_4h.bias">
 					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7739" name="Constant_299385" type="Convert" version="opset1">
+		<layer id="5174" name="__module.model.gpt_neox.layers.31.mlp.dense_h_to_4h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
 			</rt_info>
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16">
 					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7740" name="__module.model.gpt_neox.layers.31.mlp.dense_h_to_4h/aten::linear/Add" type="Add" version="opset1">
+		<layer id="5175" name="__module.model.gpt_neox.layers.31.mlp.dense_h_to_4h/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -117689,20 +77947,18 @@
 					<dim>10240</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>10240</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="5065">
+				<port id="2" precision="FP32" names="3950">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7741" name="__module.model.gpt_neox.layers.31.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
+		<layer id="5176" name="__module.model.gpt_neox.layers.31.mlp.act/aten::gelu/Gelu" type="Gelu" version="opset7">
 			<data approximation_mode="ERF" />
 			<input>
 				<port id="0" precision="FP32">
@@ -117712,15 +77968,15 @@
 				</port>
 			</input>
 			<output>
-				<port id="1" precision="FP32" names="5066">
+				<port id="1" precision="FP32" names="3951">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>10240</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7742" name="self.model.gpt_neox.layers.31.mlp.dense_4h_to_h.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 10240" offset="2621356754" size="26214400" />
+		<layer id="5177" name="self.model.gpt_neox.layers.31.mlp.dense_4h_to_h.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 10240" offset="2618304080" size="26214400" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>2560</dim>
@@ -117728,7 +77984,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7743" name="Convert_22694" type="Convert" version="opset1">
+		<layer id="5178" name="Convert_74814" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -117743,8 +77999,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7744" name="self.model.gpt_neox.layers.31.mlp.dense_4h_to_h.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="2647571154" size="2560" />
+		<layer id="5179" name="self.model.gpt_neox.layers.31.mlp.dense_4h_to_h.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="2644518480" size="2560" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>2560</dim>
@@ -117752,7 +78008,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7745" name="Convert_22697" type="Convert" version="opset1">
+		<layer id="5180" name="Convert_74817" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -117767,7 +78023,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7746" name="self.model.gpt_neox.layers.31.mlp.dense_4h_to_h.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="5181" name="self.model.gpt_neox.layers.31.mlp.dense_4h_to_h.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -117786,8 +78042,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7747" name="self.model.gpt_neox.layers.31.mlp.dense_4h_to_h.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="2647573714" size="5120" />
+		<layer id="5182" name="self.model.gpt_neox.layers.31.mlp.dense_4h_to_h.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="2644521040" size="5120" />
 			<output>
 				<port id="0" precision="FP16">
 					<dim>2560</dim>
@@ -117795,7 +78051,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7748" name="self.model.gpt_neox.layers.31.mlp.dense_4h_to_h.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="5183" name="self.model.gpt_neox.layers.31.mlp.dense_4h_to_h.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -117814,7 +78070,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7749" name="self.model.gpt_neox.layers.31.mlp.dense_4h_to_h.weight" type="Convert" version="opset1">
+		<layer id="5184" name="__module.model.gpt_neox.layers.31.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
@@ -117832,7 +78088,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7750" name="__module.model.gpt_neox.layers.31.mlp.dense_4h_to_h/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="5185" name="__module.model.gpt_neox.layers.31.mlp.dense_4h_to_h/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -117853,37 +78109,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7751" name="Constant_299386_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2647578834" size="5120" />
+		<layer id="5186" name="self.model.gpt_neox.layers.31.mlp.dense_4h_to_h.bias" type="Const" version="opset1">
+			<data element_type="bf16" shape="2560" offset="2644526160" size="5120" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16" names="self.model.gpt_neox.layers.31.mlp.dense_4h_to_h.bias">
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7752" name="Constant_299386" type="Convert" version="opset1">
+		<layer id="5187" name="__module.model.gpt_neox.layers.31.mlp.dense_4h_to_h/ov_ext::linear/ConvertLike_1" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
 			</rt_info>
 			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
+				<port id="0" precision="BF16">
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7753" name="__module.model.gpt_neox.layers.31.mlp.dense_4h_to_h/aten::linear/Add" type="Add" version="opset1">
+		<layer id="5188" name="__module.model.gpt_neox.layers.31.mlp.dense_4h_to_h/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -117892,54 +78142,42 @@
 					<dim>2560</dim>
 				</port>
 				<port id="1" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="5069,input">
+				<port id="2" precision="FP32" names="3954,input">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7754" name="Constant_280804" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="5189" name="Constant_20246" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7755" name="Constant_280807" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080218" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="5190" name="Constant_20248" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="20" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7756" name="Constant_280810" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="5191" name="Constant_20250" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7757" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Slice_6" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="5192" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Slice_6" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -117948,17 +78186,20 @@
 					<dim>240</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4947">
+				<port id="5" precision="FP32" names="3869">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -117966,15 +78207,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7758" name="Constant_18234" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="5193" name="Constant_20259" type="Const" version="opset1">
+			<data element_type="i64" shape="4" offset="201116040" size="32" />
 			<output>
-				<port id="0" precision="I64" names="4948">
+				<port id="0" precision="I64" names="3870">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7759" name="__module.model.gpt_neox.layers.31.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
+		<layer id="5194" name="__module.model.gpt_neox.layers.31.attention/aten::permute/Transpose_2" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -117987,7 +78228,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4949,query.187">
+				<port id="2" precision="FP32" names="3871,query">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -117995,41 +78236,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7760" name="Constant_280816" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="5195" name="Constant_20300" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7761" name="Constant_280819" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="5196" name="Constant_20302" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7762" name="Constant_280822" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="5197" name="Constant_20304" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7763" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Slice_7" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="5198" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Slice_7" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -118038,17 +78269,20 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4956,q">
+				<port id="5" precision="FP32" names="3878,q">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -118056,7 +78290,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7764" name="__module.model.gpt_neox.layers.31.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
+		<layer id="5199" name="__module.model.gpt_neox.layers.31.attention/aten::mul/Multiply_2" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -118073,7 +78307,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4980">
+				<port id="2" precision="FP32" names="3884">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -118081,160 +78315,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7765" name="Constant_280826" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7766" name="Constant_280825" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="5200" name="Constant_20458" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7767" name="__module.model.gpt_neox.layers.31.attention/aten::size/ShapeOf_7" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>20</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7768" name="__module.model.gpt_neox.layers.31.attention/aten::size/Constant_7" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="5201" name="Constant_20460" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="7769" name="__module.model.gpt_neox.layers.31.attention/aten::size/Gather_7" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="4981,4986" />
-			</output>
-		</layer>
-		<layer id="7770" name="__module.model.gpt_neox.layers.31.attention/aten::floor_divide/Divide_3" type="Divide" version="opset1">
-			<data auto_broadcast="numpy" m_pythondiv="true" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
-			<output>
-				<port id="2" precision="I64" />
-			</output>
-		</layer>
-		<layer id="7771" name="__module.model.gpt_neox.layers.31.attention/aten::floor_divide/Floor_3" type="Floor" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-			</input>
-			<output>
-				<port id="1" precision="I64" names="4983,4984,4988,4989" />
-			</output>
-		</layer>
-		<layer id="7772" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Constant_3" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="7773" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Unsqueeze_37" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7774" name="Constant_280824" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="5202" name="Constant_20462" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7775" name="ScatterUpdate_280827" type="ScatterUpdate" version="opset3">
-			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
 			</output>
 		</layer>
-		<layer id="7776" name="Constant_280830" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7777" name="Constant_280833" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7778" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Slice_8" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="5203" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Slice_8" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -118243,17 +78348,20 @@
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4990,x2.125">
+				<port id="5" precision="FP32" names="3894,x2.125">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -118261,32 +78369,10 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7779" name="Constant_299387_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 1, 1" offset="201162354" size="2" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7780" name="Constant_299387" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-				</port>
-			</input>
+		<layer id="5204" name="Constant_37542" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 1, 1" offset="201116148" size="4" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>1</dim>
@@ -118294,7 +78380,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7781" name="__module.model.gpt_neox.layers.31.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
+		<layer id="5205" name="__module.model.gpt_neox.layers.31.attention/aten::neg/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -118311,7 +78397,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4991">
+				<port id="2" precision="FP32" names="3895">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -118319,30 +78405,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7782" name="Constant_280878" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="5206" name="Constant_20355" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7783" name="Constant_280883" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="5207" name="Constant_20399" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116132" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+			</output>
+		</layer>
+		<layer id="5208" name="Constant_20401" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
+			<output>
+				<port id="0" precision="I64">
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7784" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Slice_9" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="5209" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Slice_9" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -118351,17 +78438,20 @@
 					<dim>20</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4985">
+				<port id="5" precision="FP32" names="3889">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -118369,7 +78459,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7785" name="__module.model.gpt_neox.layers.31.attention/aten::cat/Concat_4" type="Concat" version="opset1">
+		<layer id="5210" name="__module.model.gpt_neox.layers.31.attention/aten::cat/Concat_4" type="Concat" version="opset1">
 			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
@@ -118386,7 +78476,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4993">
+				<port id="2" precision="FP32" names="3897">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -118394,7 +78484,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7786" name="__module.model.gpt_neox.layers.31.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
+		<layer id="5211" name="__module.model.gpt_neox.layers.31.attention/aten::mul/Multiply_3" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -118411,7 +78501,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4994">
+				<port id="2" precision="FP32" names="3898">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -118419,7 +78509,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7787" name="__module.model.gpt_neox.layers.31.attention/aten::add/Add_1" type="Add" version="opset1">
+		<layer id="5212" name="__module.model.gpt_neox.layers.31.attention/aten::add/Add_1" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -118436,7 +78526,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="4995">
+				<port id="2" precision="FP32" names="3899">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -118444,41 +78534,31 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7788" name="Constant_280928" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080314" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="5213" name="Constant_20310" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116072" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7789" name="Constant_280931" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201162322" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="5214" name="Constant_20312" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116140" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7790" name="Constant_280934" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="5215" name="Constant_20314" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
 				<port id="0" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7791" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Slice_10" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
+		<layer id="5216" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Slice_10" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -118487,17 +78567,20 @@
 					<dim>80</dim>
 				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="2" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
 				</port>
 				<port id="3" precision="I64">
-					<dim>4</dim>
+					<dim>1</dim>
+				</port>
+				<port id="4" precision="I64">
+					<dim>1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="4957">
+				<port id="5" precision="FP32" names="3879">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -118505,7 +78588,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7792" name="__module.model.gpt_neox.layers.31.attention/aten::cat/Concat_5" type="Concat" version="opset1">
+		<layer id="5217" name="__module.model.gpt_neox.layers.31.attention/aten::cat/Concat_5" type="Concat" version="opset1">
 			<data axis="-1" />
 			<input>
 				<port id="0" precision="FP32">
@@ -118522,7 +78605,7 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="5013,5023,query,query.189">
+				<port id="2" precision="FP32" names="3917">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -118530,40 +78613,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7793" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Slice_11" type="Const" version="opset1">
-			<data element_type="boolean" shape="1, 1, 2048, 2048" offset="201244276" size="4194304" />
-			<output>
-				<port id="0" precision="BOOL" names="5032">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7794" name="Constant_280938" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7795" name="Constant_280937" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162298" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="5218" name="Constant_20677" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="4" size="8" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7796" name="__module.model.gpt_neox.layers.31.attention/aten::size/ShapeOf_8" type="ShapeOf" version="opset3">
+		<layer id="5219" name="ShapeOf_40913" type="ShapeOf" version="opset3">
 			<data output_type="i64" />
 			<input>
 				<port id="0" precision="FP32">
@@ -118579,386 +78637,62 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7797" name="__module.model.gpt_neox.layers.31.attention/aten::size/Constant_8" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="7798" name="__module.model.gpt_neox.layers.31.attention/aten::size/Gather_8" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="5027" />
-			</output>
-		</layer>
-		<layer id="7799" name="__module.model.gpt_neox.layers.31.attention/aten::size/ShapeOf_10" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
-					<dim>32</dim>
-					<dim>-1</dim>
-					<dim>80</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7800" name="__module.model.gpt_neox.layers.31.attention/aten::size/Constant_9" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="7801" name="__module.model.gpt_neox.layers.31.attention/aten::size/Gather_9" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64" />
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="5025" />
-			</output>
-		</layer>
-		<layer id="7802" name="__module.model.gpt_neox.layers.31.attention/aten::sub/Subtract" type="Subtract" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I64" />
-			</input>
+		<layer id="5220" name="Constant_40914" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116168" size="8" />
 			<output>
-				<port id="2" precision="I64" names="5029,5030" />
-			</output>
-		</layer>
-		<layer id="7803" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Constant_4" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="7804" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Unsqueeze_44" type="Unsqueeze" version="opset1">
-			<input>
-				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7805" name="Constant_280936" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7806" name="ScatterUpdate_280939" type="ScatterUpdate" version="opset3">
-			<input>
 				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
-				</port>
 			</output>
 		</layer>
-		<layer id="7807" name="Constant_280940" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438580" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="5221" name="Constant_40915" type="Const" version="opset1">
+			<data element_type="i64" shape="" offset="4" size="8" />
 			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7808" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Unsqueeze_45" type="Unsqueeze" version="opset1">
-			<input>
 				<port id="0" precision="I64" />
-				<port id="1" precision="I32" />
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
 			</output>
 		</layer>
-		<layer id="7809" name="ScatterUpdate_280941" type="ScatterUpdate" version="opset3">
+		<layer id="5222" name="Gather_40916" type="Gather" version="opset8">
+			<data batch_dims="0" />
 			<input>
 				<port id="0" precision="I64">
-					<dim>3</dim>
+					<dim>4</dim>
 				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
-					<dim>1</dim>
-				</port>
+				<port id="2" precision="I64" />
 			</input>
 			<output>
-				<port id="4" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7810" name="Constant_280944" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438612" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>3</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7811" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Slice_12" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 0" end_mask="1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2048</dim>
-					<dim>2048</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>3</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>3</dim>
-				</port>
 				<port id="3" precision="I64">
-					<dim>3</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="BOOL" names="5033">
-					<dim>1</dim>
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7812" name="Constant_281017" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7813" name="Constant_281018" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080186" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7814" name="Constant_281014" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="201162314" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7815" name="Constant_281013" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7816" name="ScatterUpdate_281019" type="ScatterUpdate" version="opset3">
+		<layer id="5223" name="Add_40919" type="Add" version="opset1">
+			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="3" precision="I32">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7817" name="Constant_281022" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080250" size="32" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7818" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Slice_13" type="StridedSlice" version="opset1">
-			<data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" />
-			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>2048</dim>
-				</port>
 				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="3" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="4" precision="BOOL" names="5034,5035">
-					<dim>1</dim>
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
 				</port>
-			</output>
-		</layer>
-		<layer id="7819" name="__module.model.gpt_neox.layers.31.attention/aten::where/Convert_1_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="" offset="0" size="2" />
-			<output>
-				<port id="0" precision="FP16" />
-			</output>
-		</layer>
-		<layer id="7820" name="__module.model.gpt_neox.layers.31.attention/aten::where/Convert_1" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16" />
 			</input>
 			<output>
-				<port id="1" precision="FP32" />
-			</output>
-		</layer>
-		<layer id="7821" name="__module.model.gpt_neox.layers.31.attention/aten::full/Convert" type="Const" version="opset1">
-			<data element_type="f32" shape="" offset="205438636" size="4" />
-			<output>
-				<port id="0" precision="FP32" names="5022" />
-			</output>
-		</layer>
-		<layer id="7822" name="__module.model.gpt_neox.layers.31.attention/aten::where/Select" type="Select" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="BOOL">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="1" precision="FP32" />
-				<port id="2" precision="FP32" />
-			</input>
-			<output>
-				<port id="3" precision="FP32" names="5036,causal_mask.125">
-					<dim>1</dim>
+				<port id="2" precision="I64" names="3924">
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7823" name="Add_290524" type="Add" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="1" precision="FP32">
-					<dim>-1</dim>
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>-1</dim>
-				</port>
-			</input>
+		<layer id="5224" name="Constant_20680" type="Const" version="opset1">
+			<data element_type="i64" shape="1" offset="201116024" size="8" />
 			<output>
-				<port id="2" precision="FP32">
-					<dim>-1</dim>
+				<port id="0" precision="I64">
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7824" name="ShapeOf_290525" type="ShapeOf" version="opset3">
-			<data output_type="i64" />
+		<layer id="5225" name="__module.model.gpt_neox.layers.31.attention/aten::slice/Slice_14" type="Slice" version="opset8">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -118966,56 +78700,6 @@
 					<dim>-1</dim>
 					<dim>-1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7825" name="Constant_254794" type="Const" version="opset1">
-			<data element_type="i64" shape="1" offset="2" size="8" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7826" name="__module.model.gpt_neox.layers.31.attention/aten::size/Constant_10" type="Const" version="opset1">
-			<data element_type="i32" shape="" offset="128867666" size="4" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
-			<output>
-				<port id="0" precision="I32" />
-			</output>
-		</layer>
-		<layer id="7827" name="__module.model.gpt_neox.layers.31.attention/aten::size/Gather_10" type="Gather" version="opset8">
-			<data batch_dims="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>1</dim>
-				</port>
-				<port id="2" precision="I32" />
-			</input>
-			<output>
-				<port id="3" precision="I64" names="5020">
-					<dim>1</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7828" name="__module.model.gpt_neox.layers.31.attention/prim::ListConstruct/Concat_1" type="Concat" version="opset1">
-			<data axis="0" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>1</dim>
-				</port>
 				<port id="1" precision="I64">
 					<dim>1</dim>
 				</port>
@@ -119025,64 +78709,20 @@
 				<port id="3" precision="I64">
 					<dim>1</dim>
 				</port>
-			</input>
-			<output>
-				<port id="4" precision="I64" names="5037">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7829" name="__module.model.gpt_neox.layers.31.attention/aten::expand/Abs" type="Abs" version="opset1">
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7830" name="Maximum_290526" type="Maximum" version="opset1">
-			<data auto_broadcast="numpy" />
-			<input>
-				<port id="0" precision="I64">
-					<dim>4</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
-				</port>
-			</input>
-			<output>
-				<port id="2" precision="I64">
-					<dim>4</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7831" name="__module.model.gpt_neox.layers.31.attention/aten::add/Add_2" type="Broadcast" version="opset3">
-			<data mode="numpy" />
-			<input>
-				<port id="0" precision="FP32">
-					<dim>-1</dim>
+				<port id="4" precision="I64">
 					<dim>1</dim>
-					<dim>-1</dim>
-					<dim>-1</dim>
-				</port>
-				<port id="1" precision="I64">
-					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="5039">
-					<dim>-1</dim>
+				<port id="5" precision="FP32" names="3928">
 					<dim>-1</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
 					<dim>-1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7832" name="__module.model.gpt_neox.layers.31.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
+		<layer id="5226" name="__module.model.gpt_neox.layers.31.attention/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13">
 			<data causal="false" />
 			<input>
 				<port id="0" precision="FP32">
@@ -119105,13 +78745,13 @@
 				</port>
 				<port id="3" precision="FP32">
 					<dim>-1</dim>
-					<dim>-1</dim>
+					<dim>1</dim>
 					<dim>-1</dim>
 					<dim>-1</dim>
 				</port>
 			</input>
 			<output>
-				<port id="4" precision="FP32" names="5040,5041,sdpa_result,tensor.125">
+				<port id="4" precision="FP32" names="3929,attn_output.187">
 					<dim>-1</dim>
 					<dim>32</dim>
 					<dim>-1</dim>
@@ -119119,15 +78759,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7833" name="Constant_18703" type="Const" version="opset1">
-			<data element_type="i64" shape="4" offset="201080282" size="32" />
+		<layer id="5227" name="__module.model.gpt_neox.layers.31.attention/aten::transpose/Constant" type="Const" version="opset1">
+			<data element_type="i32" shape="4" offset="201116200" size="16" />
 			<output>
-				<port id="0" precision="I64" names="5042">
+				<port id="0" precision="I32">
 					<dim>4</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7834" name="__module.model.gpt_neox.layers.31.attention/aten::permute/Transpose_3" type="Transpose" version="opset1">
+		<layer id="5228" name="__module.model.gpt_neox.layers.31.attention/aten::transpose/Transpose" type="Transpose" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -119135,12 +78775,12 @@
 					<dim>-1</dim>
 					<dim>80</dim>
 				</port>
-				<port id="1" precision="I64">
+				<port id="1" precision="I32">
 					<dim>4</dim>
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="5043">
+				<port id="2" precision="FP32" names="3930">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>32</dim>
@@ -119148,18 +78788,15 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7835" name="__module.model.gpt_neox.layers.31.attention/prim::ListConstruct/Concat_2" type="Const" version="opset1">
-			<data element_type="i64" shape="3" offset="205438650" size="24" />
-			<rt_info>
-				<attribute name="precise" version="0" />
-			</rt_info>
+		<layer id="5229" name="Constant_41015" type="Const" version="opset1">
+			<data element_type="i64" shape="3" offset="201116216" size="24" />
 			<output>
 				<port id="0" precision="I64">
 					<dim>3</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7836" name="__module.model.gpt_neox.layers.31.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
+		<layer id="5230" name="__module.model.gpt_neox.layers.31.attention/aten::view/Reshape_1" type="Reshape" version="opset1">
 			<data special_zero="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -119173,15 +78810,15 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="5048">
+				<port id="2" precision="FP32" names="3933">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7837" name="self.model.gpt_neox.layers.31.attention.dense.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 2560" offset="2647583954" size="6553600" />
+		<layer id="5231" name="self.model.gpt_neox.layers.31.attention.dense.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 2560" offset="2644531280" size="6553600" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>2560</dim>
@@ -119189,7 +78826,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7838" name="Convert_22674" type="Convert" version="opset1">
+		<layer id="5232" name="Convert_75517" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -119204,8 +78841,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7839" name="self.model.gpt_neox.layers.31.attention.dense.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="2560, 1" offset="2654137554" size="2560" />
+		<layer id="5233" name="self.model.gpt_neox.layers.31.attention.dense.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="2560, 1" offset="2651084880" size="2560" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>2560</dim>
@@ -119213,7 +78850,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7840" name="Convert_22677" type="Convert" version="opset1">
+		<layer id="5234" name="Convert_75520" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -119228,7 +78865,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7841" name="self.model.gpt_neox.layers.31.attention.dense.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="5235" name="self.model.gpt_neox.layers.31.attention.dense.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -119247,8 +78884,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7842" name="self.model.gpt_neox.layers.31.attention.dense.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="2560, 1" offset="2654140114" size="5120" />
+		<layer id="5236" name="self.model.gpt_neox.layers.31.attention.dense.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="2560, 1" offset="2651087440" size="5120" />
 			<output>
 				<port id="0" precision="FP16">
 					<dim>2560</dim>
@@ -119256,7 +78893,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7843" name="self.model.gpt_neox.layers.31.attention.dense.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="5237" name="self.model.gpt_neox.layers.31.attention.dense.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -119275,11 +78912,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7844" name="self.model.gpt_neox.layers.31.attention.dense.weight" type="Convert" version="opset1">
+		<layer id="5238" name="self.model.gpt_neox.layers.31.attention.dense.weight/fq_weights_1/convert" type="Convert" version="opset1">
 			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
 			<input>
 				<port id="0" precision="FP16">
 					<dim>2560</dim>
@@ -119293,7 +78927,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7845" name="__module.model.gpt_neox.layers.31.attention.dense/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="5239" name="__module.model.gpt_neox.layers.31.attention.dense/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -119314,37 +78948,17 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7846" name="Constant_299388_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2647578834" size="5120" />
+		<layer id="5240" name="Constant_37543" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2651092560" size="10240" />
 			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7847" name="Constant_299388" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
-			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7848" name="__module.model.gpt_neox.layers.31.attention.dense/aten::linear/Add" type="Add" version="opset1">
+		<layer id="5241" name="__module.model.gpt_neox.layers.31.attention.dense/ov_ext::linear/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -119359,14 +78973,14 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="5051,5053,input.127">
+				<port id="2" precision="FP32" names="3936,3938,input.125">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7849" name="__module.model.gpt_neox.layers.31/aten::add/Add" type="Add" version="opset1">
+		<layer id="5242" name="__module.model.gpt_neox.layers.31/aten::add/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -119381,14 +78995,14 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="5071">
+				<port id="2" precision="FP32" names="3956">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7850" name="__module.model.gpt_neox.layers.31/aten::add/Add_1" type="Add" version="opset1">
+		<layer id="5243" name="__module.model.gpt_neox.layers.31/aten::add/Add_1" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -119403,22 +79017,22 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="5072,5074">
+				<port id="2" precision="FP32" names="3957,3959">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7851" name="__module.model.gpt_neox.final_layer_norm/aten::layer_norm/Constant" type="Const" version="opset1">
-			<data element_type="i32" shape="1" offset="128867670" size="4" />
+		<layer id="5244" name="__module.model.gpt_neox.final_layer_norm/aten::layer_norm/Multiply" type="Const" version="opset1">
+			<data element_type="i32" shape="1" offset="128867668" size="4" />
 			<output>
 				<port id="0" precision="I32">
 					<dim>1</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7852" name="__module.model.gpt_neox.final_layer_norm/aten::layer_norm/MVN" type="MVN" version="opset6">
+		<layer id="5245" name="__module.model.gpt_neox.final_layer_norm/aten::layer_norm/MVN" type="MVN" version="opset6">
 			<data eps="9.9999997473787516e-06" normalize_variance="true" eps_mode="INSIDE_SQRT" />
 			<input>
 				<port id="0" precision="FP32">
@@ -119438,37 +79052,17 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7853" name="Constant_299389_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2654145234" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7854" name="Constant_299389" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="5246" name="Constant_37544" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2651102800" size="10240" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7855" name="__module.model.gpt_neox.final_layer_norm/aten::layer_norm/Multiply" type="Multiply" version="opset1">
+		<layer id="5247" name="__module.model.gpt_neox.final_layer_norm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -119490,37 +79084,17 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7856" name="Constant_299390_compressed" type="Const" version="opset1">
-			<data element_type="f16" shape="1, 1, 2560" offset="2654150354" size="5120" />
-			<output>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</output>
-		</layer>
-		<layer id="7857" name="Constant_299390" type="Convert" version="opset1">
-			<data destination_type="f32" />
-			<rt_info>
-				<attribute name="decompression" version="0" />
-			</rt_info>
-			<input>
-				<port id="0" precision="FP16">
-					<dim>1</dim>
-					<dim>1</dim>
-					<dim>2560</dim>
-				</port>
-			</input>
+		<layer id="5248" name="Constant_37545" type="Const" version="opset1">
+			<data element_type="f32" shape="1, 1, 2560" offset="2651113040" size="10240" />
 			<output>
-				<port id="1" precision="FP32">
+				<port id="0" precision="FP32">
 					<dim>1</dim>
 					<dim>1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7858" name="__module.model.gpt_neox.final_layer_norm/aten::layer_norm/Add" type="Add" version="opset1">
+		<layer id="5249" name="__module.model.gpt_neox.final_layer_norm/aten::layer_norm/Add" type="Add" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP32">
@@ -119535,15 +79109,15 @@
 				</port>
 			</input>
 			<output>
-				<port id="2" precision="FP32" names="5080,5082">
+				<port id="2" precision="FP32" names="3965,3967">
 					<dim>-1</dim>
 					<dim>-1</dim>
 					<dim>2560</dim>
 				</port>
 			</output>
 		</layer>
-		<layer id="7859" name="self.model.embed_out.weight_compressed" type="Const" version="opset1">
-			<data element_type="u8" shape="50280, 2560" offset="2654155474" size="128716800" />
+		<layer id="5250" name="self.model.embed_out.weight" type="Const" version="opset1">
+			<data element_type="u8" shape="50280, 2560" offset="2651123280" size="128716800" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>50280</dim>
@@ -119551,7 +79125,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7860" name="Convert_22704" type="Convert" version="opset1">
+		<layer id="5251" name="Convert_74174" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -119566,8 +79140,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7861" name="self.model.embed_out.weight_compressed/zero_point" type="Const" version="opset1">
-			<data element_type="u8" shape="50280, 1" offset="2782872274" size="50280" />
+		<layer id="5252" name="self.model.embed_out.weight/zero_point" type="Const" version="opset1">
+			<data element_type="u8" shape="50280, 1" offset="2779840080" size="50280" />
 			<output>
 				<port id="0" precision="U8">
 					<dim>50280</dim>
@@ -119575,7 +79149,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7862" name="Convert_22707" type="Convert" version="opset1">
+		<layer id="5253" name="Convert_74177" type="Convert" version="opset1">
 			<data destination_type="f16" />
 			<input>
 				<port id="0" precision="U8">
@@ -119590,7 +79164,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7863" name="self.model.embed_out.weight_compressed/zero_point/subtract" type="Subtract" version="opset1">
+		<layer id="5254" name="self.model.embed_out.weight/zero_point/subtract" type="Subtract" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -119609,8 +79183,8 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7864" name="self.model.embed_out.weight_compressed/scale" type="Const" version="opset1">
-			<data element_type="f16" shape="50280, 1" offset="2782922554" size="100560" />
+		<layer id="5255" name="self.model.embed_out.weight/scale" type="Const" version="opset1">
+			<data element_type="f16" shape="50280, 1" offset="2779890360" size="100560" />
 			<output>
 				<port id="0" precision="FP16">
 					<dim>50280</dim>
@@ -119618,7 +79192,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7865" name="self.model.embed_out.weight_compressed/fq_weights_1" type="Multiply" version="opset1">
+		<layer id="5256" name="self.model.embed_out.weight/fq_weights_1" type="Multiply" version="opset1">
 			<data auto_broadcast="numpy" />
 			<input>
 				<port id="0" precision="FP16">
@@ -119637,7 +79211,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7866" name="self.model.embed_out.weight" type="Convert" version="opset1">
+		<layer id="5257" name="__module.model.embed_out/ov_ext::linear/ConvertLike" type="Convert" version="opset1">
 			<data destination_type="f32" />
 			<rt_info>
 				<attribute name="decompression" version="0" />
@@ -119655,7 +79229,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7867" name="__module.model.embed_out/aten::linear/MatMul" type="MatMul" version="opset1">
+		<layer id="5258" name="__module.model.embed_out/ov_ext::linear/MatMul" type="MatMul" version="opset1">
 			<data transpose_a="false" transpose_b="true" />
 			<input>
 				<port id="0" precision="FP32">
@@ -119676,7 +79250,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7722" name="Assign_330426" type="Assign" version="opset6">
+		<layer id="5159" name="Assign_50184" type="Assign" version="opset6">
 			<data variable_id="past_key_values.0.keypresent.0.key" />
 			<input>
 				<port id="0" precision="FP32">
@@ -119695,7 +79269,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7721" name="Assign_330428" type="Assign" version="opset6">
+		<layer id="5158" name="Assign_50186" type="Assign" version="opset6">
 			<data variable_id="past_key_values.0.valuepresent.0.value" />
 			<input>
 				<port id="0" precision="FP32">
@@ -119714,7 +79288,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7720" name="Assign_330430" type="Assign" version="opset6">
+		<layer id="5157" name="Assign_50188" type="Assign" version="opset6">
 			<data variable_id="past_key_values.1.keypresent.1.key" />
 			<input>
 				<port id="0" precision="FP32">
@@ -119733,7 +79307,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7719" name="Assign_330432" type="Assign" version="opset6">
+		<layer id="5156" name="Assign_50190" type="Assign" version="opset6">
 			<data variable_id="past_key_values.1.valuepresent.1.value" />
 			<input>
 				<port id="0" precision="FP32">
@@ -119752,7 +79326,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7718" name="Assign_330434" type="Assign" version="opset6">
+		<layer id="5155" name="Assign_50192" type="Assign" version="opset6">
 			<data variable_id="past_key_values.10.keypresent.10.key" />
 			<input>
 				<port id="0" precision="FP32">
@@ -119771,7 +79345,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7717" name="Assign_330436" type="Assign" version="opset6">
+		<layer id="5154" name="Assign_50194" type="Assign" version="opset6">
 			<data variable_id="past_key_values.10.valuepresent.10.value" />
 			<input>
 				<port id="0" precision="FP32">
@@ -119790,7 +79364,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7716" name="Assign_330438" type="Assign" version="opset6">
+		<layer id="5153" name="Assign_50196" type="Assign" version="opset6">
 			<data variable_id="past_key_values.11.keypresent.11.key" />
 			<input>
 				<port id="0" precision="FP32">
@@ -119809,7 +79383,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7715" name="Assign_330440" type="Assign" version="opset6">
+		<layer id="5152" name="Assign_50198" type="Assign" version="opset6">
 			<data variable_id="past_key_values.11.valuepresent.11.value" />
 			<input>
 				<port id="0" precision="FP32">
@@ -119828,7 +79402,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7714" name="Assign_330442" type="Assign" version="opset6">
+		<layer id="5151" name="Assign_50200" type="Assign" version="opset6">
 			<data variable_id="past_key_values.12.keypresent.12.key" />
 			<input>
 				<port id="0" precision="FP32">
@@ -119847,7 +79421,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7713" name="Assign_330444" type="Assign" version="opset6">
+		<layer id="5150" name="Assign_50202" type="Assign" version="opset6">
 			<data variable_id="past_key_values.12.valuepresent.12.value" />
 			<input>
 				<port id="0" precision="FP32">
@@ -119866,7 +79440,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7712" name="Assign_330446" type="Assign" version="opset6">
+		<layer id="5149" name="Assign_50204" type="Assign" version="opset6">
 			<data variable_id="past_key_values.13.keypresent.13.key" />
 			<input>
 				<port id="0" precision="FP32">
@@ -119885,7 +79459,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7711" name="Assign_330448" type="Assign" version="opset6">
+		<layer id="5148" name="Assign_50206" type="Assign" version="opset6">
 			<data variable_id="past_key_values.13.valuepresent.13.value" />
 			<input>
 				<port id="0" precision="FP32">
@@ -119904,7 +79478,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7710" name="Assign_330450" type="Assign" version="opset6">
+		<layer id="5147" name="Assign_50208" type="Assign" version="opset6">
 			<data variable_id="past_key_values.14.keypresent.14.key" />
 			<input>
 				<port id="0" precision="FP32">
@@ -119923,7 +79497,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7709" name="Assign_330452" type="Assign" version="opset6">
+		<layer id="5146" name="Assign_50210" type="Assign" version="opset6">
 			<data variable_id="past_key_values.14.valuepresent.14.value" />
 			<input>
 				<port id="0" precision="FP32">
@@ -119942,7 +79516,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7708" name="Assign_330454" type="Assign" version="opset6">
+		<layer id="5145" name="Assign_50212" type="Assign" version="opset6">
 			<data variable_id="past_key_values.15.keypresent.15.key" />
 			<input>
 				<port id="0" precision="FP32">
@@ -119961,7 +79535,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7707" name="Assign_330456" type="Assign" version="opset6">
+		<layer id="5144" name="Assign_50214" type="Assign" version="opset6">
 			<data variable_id="past_key_values.15.valuepresent.15.value" />
 			<input>
 				<port id="0" precision="FP32">
@@ -119980,7 +79554,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7706" name="Assign_330458" type="Assign" version="opset6">
+		<layer id="5143" name="Assign_50216" type="Assign" version="opset6">
 			<data variable_id="past_key_values.16.keypresent.16.key" />
 			<input>
 				<port id="0" precision="FP32">
@@ -119999,7 +79573,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7705" name="Assign_330460" type="Assign" version="opset6">
+		<layer id="5142" name="Assign_50218" type="Assign" version="opset6">
 			<data variable_id="past_key_values.16.valuepresent.16.value" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120018,7 +79592,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7704" name="Assign_330462" type="Assign" version="opset6">
+		<layer id="5141" name="Assign_50220" type="Assign" version="opset6">
 			<data variable_id="past_key_values.17.keypresent.17.key" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120037,7 +79611,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7703" name="Assign_330464" type="Assign" version="opset6">
+		<layer id="5140" name="Assign_50222" type="Assign" version="opset6">
 			<data variable_id="past_key_values.17.valuepresent.17.value" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120056,7 +79630,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7702" name="Assign_330466" type="Assign" version="opset6">
+		<layer id="5139" name="Assign_50224" type="Assign" version="opset6">
 			<data variable_id="past_key_values.18.keypresent.18.key" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120075,7 +79649,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7701" name="Assign_330468" type="Assign" version="opset6">
+		<layer id="5138" name="Assign_50226" type="Assign" version="opset6">
 			<data variable_id="past_key_values.18.valuepresent.18.value" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120094,7 +79668,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7700" name="Assign_330470" type="Assign" version="opset6">
+		<layer id="5137" name="Assign_50228" type="Assign" version="opset6">
 			<data variable_id="past_key_values.19.keypresent.19.key" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120113,7 +79687,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7699" name="Assign_330472" type="Assign" version="opset6">
+		<layer id="5136" name="Assign_50230" type="Assign" version="opset6">
 			<data variable_id="past_key_values.19.valuepresent.19.value" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120132,7 +79706,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7698" name="Assign_330474" type="Assign" version="opset6">
+		<layer id="5135" name="Assign_50232" type="Assign" version="opset6">
 			<data variable_id="past_key_values.2.keypresent.2.key" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120151,7 +79725,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7697" name="Assign_330476" type="Assign" version="opset6">
+		<layer id="5134" name="Assign_50234" type="Assign" version="opset6">
 			<data variable_id="past_key_values.2.valuepresent.2.value" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120170,7 +79744,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7696" name="Assign_330478" type="Assign" version="opset6">
+		<layer id="5133" name="Assign_50236" type="Assign" version="opset6">
 			<data variable_id="past_key_values.20.keypresent.20.key" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120189,7 +79763,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7695" name="Assign_330480" type="Assign" version="opset6">
+		<layer id="5132" name="Assign_50238" type="Assign" version="opset6">
 			<data variable_id="past_key_values.20.valuepresent.20.value" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120208,7 +79782,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7694" name="Assign_330482" type="Assign" version="opset6">
+		<layer id="5131" name="Assign_50240" type="Assign" version="opset6">
 			<data variable_id="past_key_values.21.keypresent.21.key" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120227,7 +79801,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7693" name="Assign_330484" type="Assign" version="opset6">
+		<layer id="5130" name="Assign_50242" type="Assign" version="opset6">
 			<data variable_id="past_key_values.21.valuepresent.21.value" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120246,7 +79820,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7692" name="Assign_330486" type="Assign" version="opset6">
+		<layer id="5129" name="Assign_50244" type="Assign" version="opset6">
 			<data variable_id="past_key_values.22.keypresent.22.key" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120265,7 +79839,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7691" name="Assign_330488" type="Assign" version="opset6">
+		<layer id="5128" name="Assign_50246" type="Assign" version="opset6">
 			<data variable_id="past_key_values.22.valuepresent.22.value" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120284,7 +79858,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7690" name="Assign_330490" type="Assign" version="opset6">
+		<layer id="5127" name="Assign_50248" type="Assign" version="opset6">
 			<data variable_id="past_key_values.23.keypresent.23.key" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120303,7 +79877,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7689" name="Assign_330492" type="Assign" version="opset6">
+		<layer id="5126" name="Assign_50250" type="Assign" version="opset6">
 			<data variable_id="past_key_values.23.valuepresent.23.value" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120322,7 +79896,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7688" name="Assign_330494" type="Assign" version="opset6">
+		<layer id="5125" name="Assign_50252" type="Assign" version="opset6">
 			<data variable_id="past_key_values.24.keypresent.24.key" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120341,7 +79915,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7687" name="Assign_330496" type="Assign" version="opset6">
+		<layer id="5124" name="Assign_50254" type="Assign" version="opset6">
 			<data variable_id="past_key_values.24.valuepresent.24.value" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120360,7 +79934,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7686" name="Assign_330498" type="Assign" version="opset6">
+		<layer id="5123" name="Assign_50256" type="Assign" version="opset6">
 			<data variable_id="past_key_values.25.keypresent.25.key" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120379,7 +79953,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7685" name="Assign_330500" type="Assign" version="opset6">
+		<layer id="5122" name="Assign_50258" type="Assign" version="opset6">
 			<data variable_id="past_key_values.25.valuepresent.25.value" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120398,7 +79972,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7684" name="Assign_330502" type="Assign" version="opset6">
+		<layer id="5121" name="Assign_50260" type="Assign" version="opset6">
 			<data variable_id="past_key_values.26.keypresent.26.key" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120417,7 +79991,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7683" name="Assign_330504" type="Assign" version="opset6">
+		<layer id="5120" name="Assign_50262" type="Assign" version="opset6">
 			<data variable_id="past_key_values.26.valuepresent.26.value" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120436,7 +80010,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7682" name="Assign_330506" type="Assign" version="opset6">
+		<layer id="5119" name="Assign_50264" type="Assign" version="opset6">
 			<data variable_id="past_key_values.27.keypresent.27.key" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120455,7 +80029,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7681" name="Assign_330508" type="Assign" version="opset6">
+		<layer id="5118" name="Assign_50266" type="Assign" version="opset6">
 			<data variable_id="past_key_values.27.valuepresent.27.value" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120474,7 +80048,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7680" name="Assign_330510" type="Assign" version="opset6">
+		<layer id="5117" name="Assign_50268" type="Assign" version="opset6">
 			<data variable_id="past_key_values.28.keypresent.28.key" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120493,7 +80067,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7679" name="Assign_330512" type="Assign" version="opset6">
+		<layer id="5116" name="Assign_50270" type="Assign" version="opset6">
 			<data variable_id="past_key_values.28.valuepresent.28.value" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120512,7 +80086,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7678" name="Assign_330514" type="Assign" version="opset6">
+		<layer id="5115" name="Assign_50272" type="Assign" version="opset6">
 			<data variable_id="past_key_values.29.keypresent.29.key" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120531,7 +80105,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7677" name="Assign_330516" type="Assign" version="opset6">
+		<layer id="5114" name="Assign_50274" type="Assign" version="opset6">
 			<data variable_id="past_key_values.29.valuepresent.29.value" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120550,7 +80124,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7676" name="Assign_330518" type="Assign" version="opset6">
+		<layer id="5113" name="Assign_50276" type="Assign" version="opset6">
 			<data variable_id="past_key_values.3.keypresent.3.key" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120569,7 +80143,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7675" name="Assign_330520" type="Assign" version="opset6">
+		<layer id="5112" name="Assign_50278" type="Assign" version="opset6">
 			<data variable_id="past_key_values.3.valuepresent.3.value" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120588,7 +80162,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7674" name="Assign_330522" type="Assign" version="opset6">
+		<layer id="5111" name="Assign_50280" type="Assign" version="opset6">
 			<data variable_id="past_key_values.30.keypresent.30.key" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120607,7 +80181,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7673" name="Assign_330524" type="Assign" version="opset6">
+		<layer id="5110" name="Assign_50282" type="Assign" version="opset6">
 			<data variable_id="past_key_values.30.valuepresent.30.value" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120626,7 +80200,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7672" name="Assign_330526" type="Assign" version="opset6">
+		<layer id="5109" name="Assign_50284" type="Assign" version="opset6">
 			<data variable_id="past_key_values.31.keypresent.31.key" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120645,7 +80219,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7596" name="Assign_330528" type="Assign" version="opset6">
+		<layer id="5069" name="Assign_50286" type="Assign" version="opset6">
 			<data variable_id="past_key_values.31.valuepresent.31.value" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120664,7 +80238,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="2346" name="Assign_330530" type="Assign" version="opset6">
+		<layer id="1631" name="Assign_50288" type="Assign" version="opset6">
 			<data variable_id="past_key_values.4.keypresent.4.key" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120683,7 +80257,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="2345" name="Assign_330532" type="Assign" version="opset6">
+		<layer id="1630" name="Assign_50290" type="Assign" version="opset6">
 			<data variable_id="past_key_values.4.valuepresent.4.value" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120702,7 +80276,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="2344" name="Assign_330534" type="Assign" version="opset6">
+		<layer id="1629" name="Assign_50292" type="Assign" version="opset6">
 			<data variable_id="past_key_values.5.keypresent.5.key" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120721,7 +80295,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="2343" name="Assign_330536" type="Assign" version="opset6">
+		<layer id="1628" name="Assign_50294" type="Assign" version="opset6">
 			<data variable_id="past_key_values.5.valuepresent.5.value" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120740,7 +80314,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="2342" name="Assign_330538" type="Assign" version="opset6">
+		<layer id="1627" name="Assign_50296" type="Assign" version="opset6">
 			<data variable_id="past_key_values.6.keypresent.6.key" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120759,7 +80333,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="2341" name="Assign_330540" type="Assign" version="opset6">
+		<layer id="1626" name="Assign_50298" type="Assign" version="opset6">
 			<data variable_id="past_key_values.6.valuepresent.6.value" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120778,7 +80352,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="2340" name="Assign_330542" type="Assign" version="opset6">
+		<layer id="1625" name="Assign_50300" type="Assign" version="opset6">
 			<data variable_id="past_key_values.7.keypresent.7.key" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120797,7 +80371,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="2339" name="Assign_330544" type="Assign" version="opset6">
+		<layer id="1624" name="Assign_50302" type="Assign" version="opset6">
 			<data variable_id="past_key_values.7.valuepresent.7.value" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120816,7 +80390,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="2338" name="Assign_330546" type="Assign" version="opset6">
+		<layer id="1623" name="Assign_50304" type="Assign" version="opset6">
 			<data variable_id="past_key_values.8.keypresent.8.key" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120835,7 +80409,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="2337" name="Assign_330548" type="Assign" version="opset6">
+		<layer id="1622" name="Assign_50306" type="Assign" version="opset6">
 			<data variable_id="past_key_values.8.valuepresent.8.value" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120854,7 +80428,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="2336" name="Assign_330550" type="Assign" version="opset6">
+		<layer id="1621" name="Assign_50308" type="Assign" version="opset6">
 			<data variable_id="past_key_values.9.keypresent.9.key" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120873,7 +80447,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="2260" name="Assign_330552" type="Assign" version="opset6">
+		<layer id="1581" name="Assign_50310" type="Assign" version="opset6">
 			<data variable_id="past_key_values.9.valuepresent.9.value" />
 			<input>
 				<port id="0" precision="FP32">
@@ -120892,7 +80466,7 @@
 				</port>
 			</output>
 		</layer>
-		<layer id="7868" name="Result_139798" type="Result" version="opset1">
+		<layer id="5259" name="Result_21238" type="Result" version="opset1">
 			<input>
 				<port id="0" precision="FP32">
 					<dim>-1</dim>
@@ -120903,1533 +80477,1591 @@
 		</layer>
 	</layers>
 	<edges>
-		<edge from-layer="0" from-port="0" to-layer="3537" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="1197" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="5715" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="3387" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="2903" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="3871" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="1347" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="5807" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="1439" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="4989" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="210" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="5957" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="1589" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="4839" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="6049" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="3629" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="6199" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="1681" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="5565" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="1105" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="3295" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="5473" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="955" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="3053" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="5323" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="863" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="3145" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="5231" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="379" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="471" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="713" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="621" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="5081" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="4263" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="7606" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="6775" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="2661" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="17" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="4597" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="6925" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="4113" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="2356" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="4021" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="7167" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="7017" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="3779" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="4505" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="114" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="4355" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="2569" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="6683" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="1831" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="6291" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="1923" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="7501" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="6441" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="2073" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="5009" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="1968" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="2007" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="2126" to-port="1" />
 		<edge from-layer="0" from-port="0" to-layer="2165" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="6533" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="4747" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="7409" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="2811" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="2270" to-port="1" />
-		<edge from-layer="0" from-port="0" to-layer="7259" to-port="1" />
-		<edge from-layer="1" from-port="0" to-layer="123" to-port="0" />
-		<edge from-layer="2" from-port="0" to-layer="257" to-port="0" />
-		<edge from-layer="3" from-port="0" to-layer="251" to-port="0" />
-		<edge from-layer="3" from-port="0" to-layer="6" to-port="0" />
-		<edge from-layer="3" from-port="0" to-layer="26" to-port="0" />
-		<edge from-layer="4" from-port="0" to-layer="5" to-port="0" />
-		<edge from-layer="5" from-port="1" to-layer="14" to-port="0" />
-		<edge from-layer="6" from-port="1" to-layer="9" to-port="0" />
-		<edge from-layer="7" from-port="0" to-layer="9" to-port="1" />
-		<edge from-layer="8" from-port="0" to-layer="9" to-port="2" />
-		<edge from-layer="9" from-port="3" to-layer="7255" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="6437" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="7497" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="2899" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="6529" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="7405" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="6679" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="2807" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="6771" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="4743" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="7602" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="6921" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="3775" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="4593" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="7013" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="2657" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="7163" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="3533" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="5077" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="5227" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="5319" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="206" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="5469" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="3291" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="3049" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="5561" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="3383" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="5711" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="3141" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="5803" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="4985" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="3867" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="5953" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="3625" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="6045" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="6195" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="4835" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="6287" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="4351" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="13" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="375" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="467" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="617" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="709" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="859" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="951" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="1101" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="1193" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="1343" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="1435" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="1585" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="1677" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="1919" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="2565" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="4259" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="4109" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="110" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="2352" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="4017" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="2266" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="4501" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="2161" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="2069" to-port="0" />
-		<edge from-layer="9" from-port="3" to-layer="1827" to-port="0" />
-		<edge from-layer="10" from-port="0" to-layer="13" to-port="1" />
-		<edge from-layer="11" from-port="0" to-layer="13" to-port="2" />
-		<edge from-layer="12" from-port="0" to-layer="13" to-port="3" />
-		<edge from-layer="13" from-port="4" to-layer="14" to-port="1" />
-		<edge from-layer="14" from-port="2" to-layer="15" to-port="0" />
-		<edge from-layer="15" from-port="1" to-layer="17" to-port="0" />
-		<edge from-layer="16" from-port="0" to-layer="17" to-port="2" />
-		<edge from-layer="17" from-port="3" to-layer="2259" to-port="0" />
-		<edge from-layer="18" from-port="0" to-layer="19" to-port="0" />
-		<edge from-layer="19" from-port="1" to-layer="22" to-port="0" />
-		<edge from-layer="20" from-port="0" to-layer="21" to-port="0" />
-		<edge from-layer="21" from-port="1" to-layer="22" to-port="1" />
-		<edge from-layer="22" from-port="2" to-layer="24" to-port="0" />
-		<edge from-layer="23" from-port="0" to-layer="24" to-port="1" />
-		<edge from-layer="24" from-port="2" to-layer="25" to-port="0" />
-		<edge from-layer="25" from-port="1" to-layer="28" to-port="0" />
-		<edge from-layer="26" from-port="1" to-layer="28" to-port="1" />
-		<edge from-layer="27" from-port="0" to-layer="28" to-port="2" />
-		<edge from-layer="28" from-port="3" to-layer="294" to-port="1" />
-		<edge from-layer="28" from-port="3" to-layer="30" to-port="0" />
-		<edge from-layer="29" from-port="0" to-layer="30" to-port="1" />
-		<edge from-layer="30" from-port="2" to-layer="64" to-port="0" />
-		<edge from-layer="30" from-port="2" to-layer="33" to-port="0" />
-		<edge from-layer="31" from-port="0" to-layer="32" to-port="0" />
-		<edge from-layer="32" from-port="1" to-layer="33" to-port="1" />
-		<edge from-layer="33" from-port="2" to-layer="36" to-port="0" />
+		<edge from-layer="0" from-port="0" to-layer="2284" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="2323" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="2442" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="2481" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="2600" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="2639" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="1849" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="4970" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="4851" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="4812" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="4693" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="4654" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="4535" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="4496" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="4377" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="2758" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="4338" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="1008" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="134" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="173" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="376" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="415" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="534" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="573" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="692" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="731" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="850" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="889" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="5078" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="1047" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="1166" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="1205" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="1324" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="1363" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="1482" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="1521" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="1590" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="1640" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="1810" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="3232" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="3864" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="3113" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="3745" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="2916" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="16" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="3903" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="3074" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="3706" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="3587" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="4022" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="2955" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="2797" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="3548" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="4061" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="4219" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="3429" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="3271" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="4180" to-port="1" />
+		<edge from-layer="0" from-port="0" to-layer="3390" to-port="1" />
+		<edge from-layer="1" from-port="0" to-layer="88" to-port="0" />
+		<edge from-layer="1" from-port="0" to-layer="97" to-port="0" />
+		<edge from-layer="2" from-port="0" to-layer="253" to-port="0" />
+		<edge from-layer="2" from-port="0" to-layer="182" to-port="0" />
+		<edge from-layer="3" from-port="0" to-layer="5" to-port="0" />
+		<edge from-layer="3" from-port="0" to-layer="25" to-port="0" />
+		<edge from-layer="4" from-port="0" to-layer="13" to-port="0" />
+		<edge from-layer="5" from-port="1" to-layer="8" to-port="0" />
+		<edge from-layer="6" from-port="0" to-layer="8" to-port="1" />
+		<edge from-layer="7" from-port="0" to-layer="8" to-port="2" />
+		<edge from-layer="8" from-port="3" to-layer="1320" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="1586" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="3386" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="1517" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="169" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="372" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="411" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="3267" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="1359" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="2951" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="530" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="569" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="885" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="3228" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="3070" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="1201" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="2912" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="1162" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="688" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="727" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="1043" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="846" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="1004" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="3109" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="1964" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="4650" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="5074" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="3899" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="5005" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="4966" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="4018" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="4847" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="4808" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="4057" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="4689" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="1478" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="4531" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="2793" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="4492" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="2754" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="4176" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="4373" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="4215" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="4334" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="2280" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="3425" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="1806" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="1845" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="3544" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="2003" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="2122" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="3583" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="2161" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="1636" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="3702" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="2319" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="2438" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="3741" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="2477" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="2596" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="3860" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="2635" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="130" to-port="0" />
+		<edge from-layer="8" from-port="3" to-layer="12" to-port="0" />
+		<edge from-layer="9" from-port="0" to-layer="12" to-port="1" />
+		<edge from-layer="10" from-port="0" to-layer="12" to-port="2" />
+		<edge from-layer="11" from-port="0" to-layer="12" to-port="3" />
+		<edge from-layer="12" from-port="4" to-layer="13" to-port="1" />
+		<edge from-layer="13" from-port="2" to-layer="14" to-port="0" />
+		<edge from-layer="14" from-port="1" to-layer="16" to-port="0" />
+		<edge from-layer="15" from-port="0" to-layer="16" to-port="2" />
+		<edge from-layer="16" from-port="3" to-layer="1580" to-port="0" />
+		<edge from-layer="17" from-port="0" to-layer="18" to-port="0" />
+		<edge from-layer="18" from-port="1" to-layer="21" to-port="0" />
+		<edge from-layer="19" from-port="0" to-layer="20" to-port="0" />
+		<edge from-layer="20" from-port="1" to-layer="21" to-port="1" />
+		<edge from-layer="21" from-port="2" to-layer="23" to-port="0" />
+		<edge from-layer="22" from-port="0" to-layer="23" to-port="1" />
+		<edge from-layer="23" from-port="2" to-layer="24" to-port="0" />
+		<edge from-layer="24" from-port="1" to-layer="27" to-port="0" />
+		<edge from-layer="25" from-port="1" to-layer="27" to-port="1" />
+		<edge from-layer="25" from-port="1" to-layer="191" to-port="0" />
+		<edge from-layer="26" from-port="0" to-layer="27" to-port="2" />
+		<edge from-layer="27" from-port="3" to-layer="29" to-port="0" />
+		<edge from-layer="27" from-port="3" to-layer="290" to-port="1" />
+		<edge from-layer="28" from-port="0" to-layer="29" to-port="1" />
+		<edge from-layer="29" from-port="2" to-layer="31" to-port="0" />
+		<edge from-layer="29" from-port="2" to-layer="60" to-port="0" />
+		<edge from-layer="30" from-port="0" to-layer="31" to-port="1" />
+		<edge from-layer="31" from-port="2" to-layer="33" to-port="0" />
+		<edge from-layer="32" from-port="0" to-layer="33" to-port="1" />
+		<edge from-layer="33" from-port="2" to-layer="42" to-port="0" />
 		<edge from-layer="34" from-port="0" to-layer="35" to-port="0" />
-		<edge from-layer="35" from-port="1" to-layer="36" to-port="1" />
-		<edge from-layer="36" from-port="2" to-layer="45" to-port="0" />
-		<edge from-layer="37" from-port="0" to-layer="38" to-port="0" />
-		<edge from-layer="38" from-port="1" to-layer="41" to-port="0" />
-		<edge from-layer="39" from-port="0" to-layer="40" to-port="0" />
-		<edge from-layer="40" from-port="1" to-layer="41" to-port="1" />
-		<edge from-layer="41" from-port="2" to-layer="43" to-port="0" />
-		<edge from-layer="42" from-port="0" to-layer="43" to-port="1" />
-		<edge from-layer="43" from-port="2" to-layer="44" to-port="0" />
+		<edge from-layer="35" from-port="1" to-layer="38" to-port="0" />
+		<edge from-layer="36" from-port="0" to-layer="37" to-port="0" />
+		<edge from-layer="37" from-port="1" to-layer="38" to-port="1" />
+		<edge from-layer="38" from-port="2" to-layer="40" to-port="0" />
+		<edge from-layer="39" from-port="0" to-layer="40" to-port="1" />
+		<edge from-layer="40" from-port="2" to-layer="41" to-port="0" />
+		<edge from-layer="41" from-port="1" to-layer="42" to-port="1" />
+		<edge from-layer="42" from-port="2" to-layer="45" to-port="0" />
+		<edge from-layer="43" from-port="0" to-layer="44" to-port="0" />
 		<edge from-layer="44" from-port="1" to-layer="45" to-port="1" />
-		<edge from-layer="45" from-port="2" to-layer="48" to-port="0" />
-		<edge from-layer="46" from-port="0" to-layer="47" to-port="0" />
-		<edge from-layer="47" from-port="1" to-layer="48" to-port="1" />
-		<edge from-layer="48" from-port="2" to-layer="49" to-port="0" />
-		<edge from-layer="49" from-port="1" to-layer="58" to-port="0" />
-		<edge from-layer="50" from-port="0" to-layer="51" to-port="0" />
-		<edge from-layer="51" from-port="1" to-layer="54" to-port="0" />
-		<edge from-layer="52" from-port="0" to-layer="53" to-port="0" />
-		<edge from-layer="53" from-port="1" to-layer="54" to-port="1" />
-		<edge from-layer="54" from-port="2" to-layer="56" to-port="0" />
-		<edge from-layer="55" from-port="0" to-layer="56" to-port="1" />
-		<edge from-layer="56" from-port="2" to-layer="57" to-port="0" />
+		<edge from-layer="45" from-port="2" to-layer="46" to-port="0" />
+		<edge from-layer="46" from-port="1" to-layer="55" to-port="0" />
+		<edge from-layer="47" from-port="0" to-layer="48" to-port="0" />
+		<edge from-layer="48" from-port="1" to-layer="51" to-port="0" />
+		<edge from-layer="49" from-port="0" to-layer="50" to-port="0" />
+		<edge from-layer="50" from-port="1" to-layer="51" to-port="1" />
+		<edge from-layer="51" from-port="2" to-layer="53" to-port="0" />
+		<edge from-layer="52" from-port="0" to-layer="53" to-port="1" />
+		<edge from-layer="53" from-port="2" to-layer="54" to-port="0" />
+		<edge from-layer="54" from-port="1" to-layer="55" to-port="1" />
+		<edge from-layer="55" from-port="2" to-layer="58" to-port="0" />
+		<edge from-layer="56" from-port="0" to-layer="57" to-port="0" />
 		<edge from-layer="57" from-port="1" to-layer="58" to-port="1" />
-		<edge from-layer="58" from-port="2" to-layer="61" to-port="0" />
-		<edge from-layer="59" from-port="0" to-layer="60" to-port="0" />
-		<edge from-layer="60" from-port="1" to-layer="61" to-port="1" />
-		<edge from-layer="61" from-port="2" to-layer="293" to-port="0" />
-		<edge from-layer="62" from-port="0" to-layer="63" to-port="0" />
-		<edge from-layer="63" from-port="1" to-layer="64" to-port="1" />
-		<edge from-layer="64" from-port="2" to-layer="67" to-port="0" />
+		<edge from-layer="58" from-port="2" to-layer="289" to-port="0" />
+		<edge from-layer="59" from-port="0" to-layer="60" to-port="1" />
+		<edge from-layer="60" from-port="2" to-layer="62" to-port="0" />
+		<edge from-layer="61" from-port="0" to-layer="62" to-port="1" />
+		<edge from-layer="62" from-port="2" to-layer="71" to-port="0" />
+		<edge from-layer="63" from-port="0" to-layer="64" to-port="0" />
+		<edge from-layer="64" from-port="1" to-layer="67" to-port="0" />
 		<edge from-layer="65" from-port="0" to-layer="66" to-port="0" />
 		<edge from-layer="66" from-port="1" to-layer="67" to-port="1" />
-		<edge from-layer="67" from-port="2" to-layer="76" to-port="0" />
-		<edge from-layer="68" from-port="0" to-layer="69" to-port="0" />
-		<edge from-layer="69" from-port="1" to-layer="72" to-port="0" />
-		<edge from-layer="70" from-port="0" to-layer="71" to-port="0" />
-		<edge from-layer="71" from-port="1" to-layer="72" to-port="1" />
-		<edge from-layer="72" from-port="2" to-layer="74" to-port="0" />
-		<edge from-layer="73" from-port="0" to-layer="74" to-port="1" />
-		<edge from-layer="74" from-port="2" to-layer="75" to-port="0" />
-		<edge from-layer="75" from-port="1" to-layer="76" to-port="1" />
-		<edge from-layer="76" from-port="2" to-layer="79" to-port="0" />
-		<edge from-layer="77" from-port="0" to-layer="78" to-port="0" />
-		<edge from-layer="78" from-port="1" to-layer="79" to-port="1" />
-		<edge from-layer="79" from-port="2" to-layer="81" to-port="0" />
-		<edge from-layer="80" from-port="0" to-layer="81" to-port="1" />
-		<edge from-layer="81" from-port="2" to-layer="85" to-port="0" />
-		<edge from-layer="81" from-port="2" to-layer="98" to-port="0" />
-		<edge from-layer="81" from-port="2" to-layer="214" to-port="0" />
-		<edge from-layer="82" from-port="0" to-layer="85" to-port="1" />
-		<edge from-layer="83" from-port="0" to-layer="85" to-port="2" />
-		<edge from-layer="84" from-port="0" to-layer="85" to-port="3" />
-		<edge from-layer="85" from-port="4" to-layer="87" to-port="0" />
-		<edge from-layer="86" from-port="0" to-layer="87" to-port="1" />
-		<edge from-layer="87" from-port="2" to-layer="91" to-port="0" />
-		<edge from-layer="87" from-port="2" to-layer="165" to-port="0" />
-		<edge from-layer="88" from-port="0" to-layer="91" to-port="1" />
-		<edge from-layer="89" from-port="0" to-layer="91" to-port="2" />
-		<edge from-layer="90" from-port="0" to-layer="91" to-port="3" />
-		<edge from-layer="91" from-port="4" to-layer="131" to-port="0" />
-		<edge from-layer="91" from-port="4" to-layer="128" to-port="0" />
-		<edge from-layer="91" from-port="4" to-layer="144" to-port="0" />
-		<edge from-layer="91" from-port="4" to-layer="150" to-port="0" />
-		<edge from-layer="92" from-port="0" to-layer="93" to-port="0" />
-		<edge from-layer="93" from-port="1" to-layer="122" to-port="0" />
-		<edge from-layer="94" from-port="0" to-layer="122" to-port="1" />
-		<edge from-layer="95" from-port="0" to-layer="98" to-port="1" />
-		<edge from-layer="96" from-port="0" to-layer="98" to-port="2" />
-		<edge from-layer="97" from-port="0" to-layer="98" to-port="3" />
-		<edge from-layer="98" from-port="4" to-layer="100" to-port="0" />
-		<edge from-layer="99" from-port="0" to-layer="100" to-port="1" />
-		<edge from-layer="100" from-port="2" to-layer="101" to-port="0" />
-		<edge from-layer="100" from-port="2" to-layer="198" to-port="0" />
-		<edge from-layer="100" from-port="2" to-layer="170" to-port="0" />
-		<edge from-layer="101" from-port="1" to-layer="104" to-port="0" />
-		<edge from-layer="102" from-port="0" to-layer="4011" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="1579" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="866" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="853" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="1095" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="624" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="611" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="1108" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="1834" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="382" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="1337" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="1821" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="369" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="1350" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="1592" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="7157" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="6431" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="3298" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="5705" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="3527" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="5718" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="5947" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="3540" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="5960" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="6189" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="7625" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="7622" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="6202" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="5476" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="6444" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="7412" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="6673" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="7399" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="6686" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="6915" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="3769" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="7170" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="6928" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="3782" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="2801" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="2076" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="2286" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="2289" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="4024" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="4253" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="2559" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="4266" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="4495" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="2572" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="4508" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="4737" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="2063" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="2814" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="4750" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="4979" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="3043" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="4992" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="3056" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="5221" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="5234" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="5463" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="3285" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="259" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="117" to-port="1" />
-		<edge from-layer="102" from-port="0" to-layer="104" to-port="1" />
-		<edge from-layer="103" from-port="0" to-layer="104" to-port="2" />
-		<edge from-layer="104" from-port="3" to-layer="118" to-port="0" />
-		<edge from-layer="105" from-port="0" to-layer="106" to-port="0" />
-		<edge from-layer="106" from-port="1" to-layer="111" to-port="0" />
-		<edge from-layer="107" from-port="0" to-layer="110" to-port="1" />
-		<edge from-layer="108" from-port="0" to-layer="110" to-port="2" />
-		<edge from-layer="109" from-port="0" to-layer="110" to-port="3" />
-		<edge from-layer="110" from-port="4" to-layer="111" to-port="1" />
-		<edge from-layer="111" from-port="2" to-layer="112" to-port="0" />
-		<edge from-layer="112" from-port="1" to-layer="114" to-port="0" />
-		<edge from-layer="113" from-port="0" to-layer="114" to-port="2" />
-		<edge from-layer="114" from-port="3" to-layer="115" to-port="0" />
-		<edge from-layer="114" from-port="3" to-layer="200" to-port="0" />
-		<edge from-layer="115" from-port="1" to-layer="117" to-port="0" />
-		<edge from-layer="116" from-port="0" to-layer="117" to-port="2" />
-		<edge from-layer="117" from-port="3" to-layer="118" to-port="1" />
-		<edge from-layer="118" from-port="2" to-layer="120" to-port="0" />
-		<edge from-layer="119" from-port="0" to-layer="120" to-port="1" />
-		<edge from-layer="120" from-port="2" to-layer="156" to-port="2" />
-		<edge from-layer="120" from-port="2" to-layer="122" to-port="2" />
-		<edge from-layer="121" from-port="0" to-layer="122" to-port="3" />
-		<edge from-layer="122" from-port="4" to-layer="125" to-port="0" />
-		<edge from-layer="123" from-port="1" to-layer="4031" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="6935" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="6965" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="3789" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="4061" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="7177" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="2113" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="7207" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="419" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="4303" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="7419" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="7449" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="2083" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="7632" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="389" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="7662" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="1871" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="1841" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="1629" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="3577" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="3547" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="1599" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="3335" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="2296" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="4999" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="5029" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="631" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="125" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="5241" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="5271" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="4787" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="5483" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="5513" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="4757" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="5725" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="5755" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="4545" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="4515" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="6723" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="5967" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="5997" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="661" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="158" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="6209" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="6239" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="2326" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="873" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="6451" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="6481" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="3819" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="4273" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="6693" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="2579" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="3063" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="3093" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="1115" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="1387" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="2821" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="2609" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="3305" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="1357" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="903" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="1145" to-port="1" />
-		<edge from-layer="123" from-port="1" to-layer="2851" to-port="1" />
-		<edge from-layer="124" from-port="0" to-layer="125" to-port="2" />
-		<edge from-layer="125" from-port="3" to-layer="127" to-port="0" />
-		<edge from-layer="126" from-port="0" to-layer="5756" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="4516" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="5968" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="4546" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="7178" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="390" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="904" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="7208" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="420" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="2580" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="2084" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="3094" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="5726" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="4758" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="2610" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="5484" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="258" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="5514" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="7420" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="7450" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="4062" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="2852" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="6482" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="3820" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="6452" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="4274" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="1146" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="4304" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="2327" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="6694" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="2822" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="6724" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="6240" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="3064" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="6210" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="3790" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="159" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="662" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="1116" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="6936" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="6966" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="5998" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="2114" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="4032" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="2297" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="1842" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="874" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="1600" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="3578" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="5272" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="1872" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="5030" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="632" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="5000" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="5242" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="4788" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="7663" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="1358" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="1388" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="7633" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="3306" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="1630" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="127" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="3336" to-port="1" />
-		<edge from-layer="126" from-port="0" to-layer="3548" to-port="1" />
-		<edge from-layer="127" from-port="2" to-layer="171" to-port="1" />
-		<edge from-layer="127" from-port="2" to-layer="128" to-port="1" />
-		<edge from-layer="128" from-port="2" to-layer="161" to-port="0" />
-		<edge from-layer="129" from-port="0" to-layer="141" to-port="0" />
-		<edge from-layer="130" from-port="0" to-layer="141" to-port="1" />
-		<edge from-layer="131" from-port="1" to-layer="134" to-port="0" />
-		<edge from-layer="132" from-port="0" to-layer="3554" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="176" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="1122" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="6983" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="3595" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="6942" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="6015" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="5047" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="880" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="1606" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="5974" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="4038" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="2403" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="2828" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="3312" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="4280" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="2869" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="6499" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="679" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="6458" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="3353" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="1405" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="5006" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="4321" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="134" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="1163" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="6700" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="6257" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="2303" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="6741" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="3837" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="6216" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="3796" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="4079" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="2586" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="4563" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="3111" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="2131" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="5732" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="1848" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="2627" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="1889" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="7426" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="2090" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="437" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="5289" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="7467" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="5531" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="7769" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="5490" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="7639" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="396" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="4764" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="1364" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="7225" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="4805" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="1647" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="5248" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="5773" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="3070" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="921" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="638" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="4522" to-port="1" />
-		<edge from-layer="132" from-port="0" to-layer="7184" to-port="1" />
+		<edge from-layer="67" from-port="2" to-layer="69" to-port="0" />
+		<edge from-layer="68" from-port="0" to-layer="69" to-port="1" />
+		<edge from-layer="69" from-port="2" to-layer="70" to-port="0" />
+		<edge from-layer="70" from-port="1" to-layer="71" to-port="1" />
+		<edge from-layer="71" from-port="2" to-layer="73" to-port="0" />
+		<edge from-layer="72" from-port="0" to-layer="73" to-port="1" />
+		<edge from-layer="73" from-port="2" to-layer="75" to-port="0" />
+		<edge from-layer="74" from-port="0" to-layer="75" to-port="1" />
+		<edge from-layer="75" from-port="2" to-layer="138" to-port="0" />
+		<edge from-layer="75" from-port="2" to-layer="80" to-port="0" />
+		<edge from-layer="75" from-port="2" to-layer="177" to-port="0" />
+		<edge from-layer="76" from-port="0" to-layer="80" to-port="1" />
+		<edge from-layer="77" from-port="0" to-layer="80" to-port="2" />
+		<edge from-layer="78" from-port="0" to-layer="80" to-port="3" />
+		<edge from-layer="79" from-port="0" to-layer="2566" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2425" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2572" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="430" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2577" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2583" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2470" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2496" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2590" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2485" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2312" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2136" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2169" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2180" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2250" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2256" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2261" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2267" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2274" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2288" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2294" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2299" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2305" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2463" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2327" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="675" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2338" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2408" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2414" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2419" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2154" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2432" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="80" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2446" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2452" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2457" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4855" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4985" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="391" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4980" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4974" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4960" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="386" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4953" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4947" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4942" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4936" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="380" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4866" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4991" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4840" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="664" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4833" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4827" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4822" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4816" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4802" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="366" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4795" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4789" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4784" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4778" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2643" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2604" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2610" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2615" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2621" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="5225" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2628" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="419" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="5216" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="5209" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="5203" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="5198" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="5192" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="669" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2654" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="5106" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="5099" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="5093" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="404" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="5088" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="5082" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="5065" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="5024" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="5013" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="397" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4998" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="833" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1296" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1301" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="854" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="840" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1307" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="577" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1314" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1328" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1334" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1339" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1345" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1352" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1367" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1378" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1448" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1290" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1454" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="562" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1459" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="827" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="822" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1465" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1472" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1486" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1492" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="555" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1497" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1503" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1510" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="816" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="746" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1132" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="980" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="985" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="991" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="998" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="904" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1012" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="893" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1018" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1023" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1029" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="878" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1036" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="109" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1051" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1062" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="549" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1138" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1143" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1149" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1156" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="871" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="865" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="860" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1170" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1176" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="588" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1181" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1187" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1194" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1209" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1220" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1978" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1831" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1838" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="511" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1853" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1864" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="696" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="682" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1934" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1940" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="506" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1945" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1951" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1958" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="500" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1972" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1825" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="658" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1983" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1989" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1996" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2011" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2022" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2092" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2098" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2103" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2109" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2116" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="86" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2130" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="974" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2141" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1684" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1525" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1536" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1577" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="735" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="544" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1594" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1600" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1605" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="115" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="720" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1611" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="538" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1618" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1673" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1679" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2147" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1690" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="713" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="707" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1697" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1706" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1776" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1782" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="524" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1787" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1793" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1800" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1814" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="702" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="1820" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="517" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3672" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3836" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3830" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3760" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3749" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3734" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3727" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3721" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3716" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3710" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3696" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3689" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3683" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3678" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3841" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3602" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3591" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3576" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3569" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3563" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3558" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3552" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3538" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3531" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3525" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3520" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3514" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3444" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3988" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4146" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4076" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4065" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4050" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4043" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4037" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2801" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4032" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4026" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4012" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4005" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3999" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3994" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3433" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2812" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="252" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3918" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3907" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3892" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="247" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3885" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3879" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3874" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3868" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3854" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3847" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3078" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="162" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3204" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3198" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3128" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3117" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2920" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="155" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3102" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3095" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3089" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2926" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="149" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3084" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3209" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="144" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3064" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2931" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3057" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3051" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="138" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3046" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3040" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2937" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2970" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2959" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2944" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3286" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3418" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3411" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2882" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3405" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3400" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3394" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2888" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3380" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3373" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3367" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2893" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3362" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3356" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4708" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3275" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2899" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3260" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="177" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3253" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3247" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="124" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3242" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3236" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2906" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3222" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="3215" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4310" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4342" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4328" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4550" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4321" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2768" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2730" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4620" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4315" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="342" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4631" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4539" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4304" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4637" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4234" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2773" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2724" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="348" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4644" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4223" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4208" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2779" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4381" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4479" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4473" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4486" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2748" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4468" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4462" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4500" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4392" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4506" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2741" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4626" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4366" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4359" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4511" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4353" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4517" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4524" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4348" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2735" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2762" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4669" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4152" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4157" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="272" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4697" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4163" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="359" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4682" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4675" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="2786" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4170" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="353" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4184" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4190" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4664" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4658" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4201" to-port="4" />
+		<edge from-layer="79" from-port="0" to-layer="4195" to-port="4" />
+		<edge from-layer="80" from-port="5" to-layer="82" to-port="0" />
+		<edge from-layer="81" from-port="0" to-layer="82" to-port="1" />
+		<edge from-layer="82" from-port="2" to-layer="86" to-port="0" />
+		<edge from-layer="82" from-port="2" to-layer="124" to-port="0" />
+		<edge from-layer="83" from-port="0" to-layer="86" to-port="1" />
+		<edge from-layer="84" from-port="0" to-layer="86" to-port="2" />
+		<edge from-layer="85" from-port="0" to-layer="86" to-port="3" />
+		<edge from-layer="86" from-port="5" to-layer="105" to-port="0" />
+		<edge from-layer="86" from-port="5" to-layer="109" to-port="0" />
+		<edge from-layer="86" from-port="5" to-layer="115" to-port="0" />
+		<edge from-layer="87" from-port="0" to-layer="95" to-port="0" />
+		<edge from-layer="88" from-port="1" to-layer="91" to-port="0" />
+		<edge from-layer="89" from-port="0" to-layer="91" to-port="1" />
+		<edge from-layer="90" from-port="0" to-layer="91" to-port="2" />
+		<edge from-layer="91" from-port="3" to-layer="94" to-port="0" />
+		<edge from-layer="92" from-port="0" to-layer="94" to-port="1" />
+		<edge from-layer="93" from-port="0" to-layer="94" to-port="2" />
+		<edge from-layer="94" from-port="3" to-layer="95" to-port="1" />
+		<edge from-layer="95" from-port="2" to-layer="99" to-port="0" />
+		<edge from-layer="96" from-port="0" to-layer="118" to-port="1" />
+		<edge from-layer="96" from-port="0" to-layer="104" to-port="1" />
+		<edge from-layer="96" from-port="0" to-layer="253" to-port="1" />
+		<edge from-layer="96" from-port="0" to-layer="228" to-port="1" />
+		<edge from-layer="96" from-port="0" to-layer="97" to-port="1" />
+		<edge from-layer="97" from-port="2" to-layer="98" to-port="0" />
+		<edge from-layer="98" from-port="1" to-layer="99" to-port="1" />
+		<edge from-layer="99" from-port="2" to-layer="101" to-port="0" />
+		<edge from-layer="100" from-port="0" to-layer="101" to-port="1" />
+		<edge from-layer="101" from-port="2" to-layer="102" to-port="0" />
+		<edge from-layer="101" from-port="2" to-layer="102" to-port="1" />
+		<edge from-layer="102" from-port="2" to-layer="117" to-port="0" />
+		<edge from-layer="102" from-port="2" to-layer="103" to-port="0" />
+		<edge from-layer="103" from-port="1" to-layer="104" to-port="0" />
+		<edge from-layer="104" from-port="2" to-layer="349" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="665" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="1493" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="3085" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="3401" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="1455" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="703" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="4785" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="2889" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="3243" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="2927" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="4627" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="3205" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="4943" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="2415" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="861" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="1139" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="4823" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="145" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="1177" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="387" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="3363" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="4665" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="2731" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="105" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="981" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="1019" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="1297" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="823" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="4981" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="1335" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="3047" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="4507" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="4349" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="4191" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="507" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="1680" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="3837" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="3995" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="4033" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="2453" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="5199" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="2099" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="4311" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="2257" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="1941" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="3679" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="1783" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="2611" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="2769" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="545" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="4469" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="5089" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="2137" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="4153" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="1979" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="3521" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="3717" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="3875" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="2295" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="1601" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="1821" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="2573" to-port="1" />
+		<edge from-layer="104" from-port="2" to-layer="3559" to-port="1" />
+		<edge from-layer="105" from-port="2" to-layer="120" to-port="0" />
+		<edge from-layer="106" from-port="0" to-layer="109" to-port="1" />
+		<edge from-layer="107" from-port="0" to-layer="109" to-port="2" />
+		<edge from-layer="108" from-port="0" to-layer="109" to-port="3" />
+		<edge from-layer="109" from-port="5" to-layer="111" to-port="0" />
+		<edge from-layer="110" from-port="0" to-layer="111" to-port="1" />
+		<edge from-layer="111" from-port="2" to-layer="116" to-port="0" />
+		<edge from-layer="112" from-port="0" to-layer="115" to-port="1" />
+		<edge from-layer="113" from-port="0" to-layer="115" to-port="2" />
+		<edge from-layer="114" from-port="0" to-layer="115" to-port="3" />
+		<edge from-layer="115" from-port="5" to-layer="116" to-port="1" />
+		<edge from-layer="116" from-port="2" to-layer="119" to-port="0" />
+		<edge from-layer="117" from-port="1" to-layer="118" to-port="0" />
+		<edge from-layer="118" from-port="2" to-layer="3887" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="3097" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="993" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="4203" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="1991" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="835" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="4835" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="4677" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="361" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="2427" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="677" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="2939" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="4045" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="2307" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="4797" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="715" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="4165" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="2111" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="3059" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="2269" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="4007" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="4481" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="2465" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="2149" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="2781" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="873" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="1613" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="2901" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="119" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="1309" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="4519" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="1692" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="4993" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="3571" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="1347" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="4361" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="157" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="3533" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="3375" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="5101" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="2743" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="399" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="1467" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="557" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="1505" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="3413" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="1833" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="4955" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="3217" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="1151" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="1953" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="1031" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="3255" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="2623" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="3849" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="3729" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="4323" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="1189" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="4639" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="519" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="2585" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="3691" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="1795" to-port="1" />
+		<edge from-layer="118" from-port="2" to-layer="5211" to-port="1" />
+		<edge from-layer="119" from-port="2" to-layer="120" to-port="1" />
+		<edge from-layer="120" from-port="2" to-layer="125" to-port="0" />
+		<edge from-layer="121" from-port="0" to-layer="124" to-port="1" />
+		<edge from-layer="122" from-port="0" to-layer="124" to-port="2" />
+		<edge from-layer="123" from-port="0" to-layer="124" to-port="3" />
+		<edge from-layer="124" from-port="5" to-layer="125" to-port="1" />
+		<edge from-layer="125" from-port="2" to-layer="273" to-port="0" />
+		<edge from-layer="126" from-port="0" to-layer="131" to-port="0" />
+		<edge from-layer="127" from-port="0" to-layer="130" to-port="1" />
+		<edge from-layer="128" from-port="0" to-layer="130" to-port="2" />
+		<edge from-layer="129" from-port="0" to-layer="130" to-port="3" />
+		<edge from-layer="130" from-port="4" to-layer="131" to-port="1" />
+		<edge from-layer="131" from-port="2" to-layer="132" to-port="0" />
+		<edge from-layer="132" from-port="1" to-layer="134" to-port="0" />
 		<edge from-layer="133" from-port="0" to-layer="134" to-port="2" />
-		<edge from-layer="134" from-port="3" to-layer="136" to-port="0" />
-		<edge from-layer="135" from-port="0" to-layer="6217" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="4806" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="6459" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="4765" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="2628" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="5249" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="5975" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="639" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="2304" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="5491" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="3838" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="922" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="5290" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="4281" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="2829" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="5048" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="5532" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="1123" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="6016" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="4322" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="5733" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="5007" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="4523" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="2587" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="4564" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="5774" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="6258" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="136" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="881" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="3313" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="7468" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="6943" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="6984" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="177" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="4039" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="438" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="2404" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="3071" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="7185" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="7226" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="2132" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="3112" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="2091" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="7427" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="6500" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="397" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="7640" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="1365" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="7770" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="1890" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="1849" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="1648" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="1607" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="3596" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="3555" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="1406" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="3354" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="6701" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="6742" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="1164" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="3797" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="2870" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="4080" to-port="1" />
-		<edge from-layer="135" from-port="0" to-layer="680" to-port="1" />
-		<edge from-layer="136" from-port="2" to-layer="137" to-port="0" />
-		<edge from-layer="137" from-port="1" to-layer="139" to-port="0" />
-		<edge from-layer="138" from-port="0" to-layer="139" to-port="1" />
-		<edge from-layer="139" from-port="2" to-layer="141" to-port="2" />
-		<edge from-layer="140" from-port="0" to-layer="141" to-port="3" />
-		<edge from-layer="141" from-port="4" to-layer="150" to-port="2" />
-		<edge from-layer="141" from-port="4" to-layer="144" to-port="1" />
+		<edge from-layer="134" from-port="3" to-layer="214" to-port="0" />
+		<edge from-layer="134" from-port="3" to-layer="164" to-port="0" />
+		<edge from-layer="135" from-port="0" to-layer="138" to-port="1" />
+		<edge from-layer="136" from-port="0" to-layer="138" to-port="2" />
+		<edge from-layer="137" from-port="0" to-layer="138" to-port="3" />
+		<edge from-layer="138" from-port="5" to-layer="140" to-port="0" />
+		<edge from-layer="139" from-port="0" to-layer="140" to-port="1" />
+		<edge from-layer="140" from-port="2" to-layer="144" to-port="0" />
+		<edge from-layer="140" from-port="2" to-layer="162" to-port="0" />
+		<edge from-layer="141" from-port="0" to-layer="144" to-port="1" />
 		<edge from-layer="142" from-port="0" to-layer="144" to-port="2" />
 		<edge from-layer="143" from-port="0" to-layer="144" to-port="3" />
-		<edge from-layer="144" from-port="4" to-layer="147" to-port="0" />
-		<edge from-layer="145" from-port="0" to-layer="146" to-port="0" />
-		<edge from-layer="146" from-port="1" to-layer="147" to-port="1" />
-		<edge from-layer="147" from-port="2" to-layer="151" to-port="0" />
-		<edge from-layer="148" from-port="0" to-layer="150" to-port="1" />
-		<edge from-layer="149" from-port="0" to-layer="150" to-port="3" />
-		<edge from-layer="150" from-port="4" to-layer="151" to-port="1" />
-		<edge from-layer="151" from-port="2" to-layer="160" to-port="0" />
-		<edge from-layer="152" from-port="0" to-layer="153" to-port="0" />
-		<edge from-layer="153" from-port="1" to-layer="156" to-port="0" />
-		<edge from-layer="154" from-port="0" to-layer="156" to-port="1" />
-		<edge from-layer="155" from-port="0" to-layer="156" to-port="3" />
-		<edge from-layer="156" from-port="4" to-layer="158" to-port="0" />
-		<edge from-layer="157" from-port="0" to-layer="158" to-port="2" />
-		<edge from-layer="158" from-port="3" to-layer="159" to-port="0" />
-		<edge from-layer="159" from-port="2" to-layer="193" to-port="1" />
-		<edge from-layer="159" from-port="2" to-layer="160" to-port="1" />
-		<edge from-layer="160" from-port="2" to-layer="161" to-port="1" />
-		<edge from-layer="161" from-port="2" to-layer="166" to-port="0" />
-		<edge from-layer="162" from-port="0" to-layer="165" to-port="1" />
-		<edge from-layer="163" from-port="0" to-layer="165" to-port="2" />
-		<edge from-layer="164" from-port="0" to-layer="165" to-port="3" />
-		<edge from-layer="165" from-port="4" to-layer="166" to-port="1" />
-		<edge from-layer="166" from-port="2" to-layer="225" to-port="0" />
-		<edge from-layer="166" from-port="2" to-layer="276" to-port="0" />
-		<edge from-layer="167" from-port="0" to-layer="170" to-port="1" />
-		<edge from-layer="168" from-port="0" to-layer="170" to-port="2" />
-		<edge from-layer="169" from-port="0" to-layer="170" to-port="3" />
-		<edge from-layer="170" from-port="4" to-layer="191" to-port="0" />
-		<edge from-layer="170" from-port="4" to-layer="174" to-port="0" />
-		<edge from-layer="170" from-port="4" to-layer="185" to-port="0" />
-		<edge from-layer="170" from-port="4" to-layer="171" to-port="0" />
-		<edge from-layer="171" from-port="2" to-layer="194" to-port="0" />
-		<edge from-layer="172" from-port="0" to-layer="182" to-port="0" />
-		<edge from-layer="173" from-port="0" to-layer="182" to-port="1" />
-		<edge from-layer="174" from-port="1" to-layer="176" to-port="0" />
-		<edge from-layer="175" from-port="0" to-layer="176" to-port="2" />
-		<edge from-layer="176" from-port="3" to-layer="177" to-port="0" />
-		<edge from-layer="177" from-port="2" to-layer="178" to-port="0" />
-		<edge from-layer="178" from-port="1" to-layer="180" to-port="0" />
-		<edge from-layer="179" from-port="0" to-layer="180" to-port="1" />
-		<edge from-layer="180" from-port="2" to-layer="182" to-port="2" />
-		<edge from-layer="181" from-port="0" to-layer="182" to-port="3" />
-		<edge from-layer="182" from-port="4" to-layer="185" to-port="1" />
-		<edge from-layer="182" from-port="4" to-layer="191" to-port="2" />
-		<edge from-layer="183" from-port="0" to-layer="185" to-port="2" />
-		<edge from-layer="184" from-port="0" to-layer="185" to-port="3" />
-		<edge from-layer="185" from-port="4" to-layer="188" to-port="0" />
-		<edge from-layer="186" from-port="0" to-layer="187" to-port="0" />
-		<edge from-layer="187" from-port="1" to-layer="188" to-port="1" />
-		<edge from-layer="188" from-port="2" to-layer="192" to-port="0" />
-		<edge from-layer="189" from-port="0" to-layer="191" to-port="1" />
-		<edge from-layer="190" from-port="0" to-layer="191" to-port="3" />
-		<edge from-layer="191" from-port="4" to-layer="192" to-port="1" />
-		<edge from-layer="192" from-port="2" to-layer="193" to-port="0" />
-		<edge from-layer="193" from-port="2" to-layer="194" to-port="1" />
-		<edge from-layer="194" from-port="2" to-layer="199" to-port="0" />
-		<edge from-layer="195" from-port="0" to-layer="198" to-port="1" />
-		<edge from-layer="196" from-port="0" to-layer="198" to-port="2" />
-		<edge from-layer="197" from-port="0" to-layer="198" to-port="3" />
-		<edge from-layer="198" from-port="4" to-layer="199" to-port="1" />
-		<edge from-layer="199" from-port="2" to-layer="200" to-port="1" />
-		<edge from-layer="200" from-port="2" to-layer="7722" to-port="0" />
-		<edge from-layer="200" from-port="2" to-layer="276" to-port="1" />
-		<edge from-layer="200" from-port="2" to-layer="221" to-port="0" />
-		<edge from-layer="201" from-port="0" to-layer="202" to-port="0" />
-		<edge from-layer="202" from-port="1" to-layer="207" to-port="0" />
-		<edge from-layer="203" from-port="0" to-layer="206" to-port="1" />
-		<edge from-layer="204" from-port="0" to-layer="206" to-port="2" />
-		<edge from-layer="205" from-port="0" to-layer="206" to-port="3" />
-		<edge from-layer="206" from-port="4" to-layer="207" to-port="1" />
-		<edge from-layer="207" from-port="2" to-layer="208" to-port="0" />
-		<edge from-layer="208" from-port="1" to-layer="210" to-port="0" />
-		<edge from-layer="209" from-port="0" to-layer="210" to-port="2" />
-		<edge from-layer="210" from-port="3" to-layer="217" to-port="0" />
-		<edge from-layer="211" from-port="0" to-layer="214" to-port="1" />
-		<edge from-layer="212" from-port="0" to-layer="214" to-port="2" />
-		<edge from-layer="213" from-port="0" to-layer="214" to-port="3" />
-		<edge from-layer="214" from-port="4" to-layer="216" to-port="0" />
-		<edge from-layer="215" from-port="0" to-layer="216" to-port="1" />
-		<edge from-layer="216" from-port="2" to-layer="217" to-port="1" />
-		<edge from-layer="217" from-port="2" to-layer="7721" to-port="0" />
-		<edge from-layer="217" from-port="2" to-layer="276" to-port="2" />
-		<edge from-layer="218" from-port="0" to-layer="237" to-port="0" />
-		<edge from-layer="219" from-port="0" to-layer="232" to-port="0" />
-		<edge from-layer="220" from-port="0" to-layer="235" to-port="1" />
-		<edge from-layer="220" from-port="0" to-layer="232" to-port="1" />
-		<edge from-layer="221" from-port="1" to-layer="224" to-port="0" />
-		<edge from-layer="222" from-port="0" to-layer="4371" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="224" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="971" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="968" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="4613" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="1455" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="1452" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="4610" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="4855" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="1210" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="4852" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="4368" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="2178" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="1694" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="2181" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="1697" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="729" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="726" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="4129" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="1939" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="4126" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="1936" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="1213" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="6304" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="7033" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="7030" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="6788" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="6549" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="6546" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="484" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="487" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="6307" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="7272" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="6065" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="6062" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="5823" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="5820" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="227" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="3884" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="2677" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="6791" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="7275" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="2432" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="2435" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="7514" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="2916" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="7517" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="2919" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="7798" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="7801" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="3645" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="3642" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="3403" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="3400" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="3161" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="3158" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="5336" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="2674" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="5339" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="3887" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="5578" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="5094" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="5581" to-port="1" />
-		<edge from-layer="222" from-port="0" to-layer="5097" to-port="1" />
-		<edge from-layer="223" from-port="0" to-layer="224" to-port="2" />
-		<edge from-layer="224" from-port="3" to-layer="228" to-port="0" />
-		<edge from-layer="224" from-port="3" to-layer="234" to-port="0" />
-		<edge from-layer="225" from-port="1" to-layer="271" to-port="0" />
-		<edge from-layer="225" from-port="1" to-layer="227" to-port="0" />
-		<edge from-layer="226" from-port="0" to-layer="227" to-port="2" />
-		<edge from-layer="227" from-port="3" to-layer="228" to-port="1" />
-		<edge from-layer="228" from-port="2" to-layer="230" to-port="0" />
-		<edge from-layer="229" from-port="0" to-layer="234" to-port="1" />
-		<edge from-layer="229" from-port="0" to-layer="230" to-port="1" />
-		<edge from-layer="230" from-port="2" to-layer="232" to-port="2" />
-		<edge from-layer="231" from-port="0" to-layer="232" to-port="3" />
-		<edge from-layer="231" from-port="0" to-layer="235" to-port="3" />
-		<edge from-layer="232" from-port="4" to-layer="237" to-port="1" />
-		<edge from-layer="233" from-port="0" to-layer="235" to-port="0" />
-		<edge from-layer="234" from-port="2" to-layer="242" to-port="2" />
-		<edge from-layer="234" from-port="2" to-layer="235" to-port="2" />
-		<edge from-layer="235" from-port="4" to-layer="237" to-port="2" />
-		<edge from-layer="236" from-port="0" to-layer="237" to-port="3" />
-		<edge from-layer="237" from-port="4" to-layer="244" to-port="0" />
-		<edge from-layer="238" from-port="0" to-layer="244" to-port="1" />
-		<edge from-layer="239" from-port="0" to-layer="242" to-port="0" />
-		<edge from-layer="240" from-port="0" to-layer="242" to-port="1" />
-		<edge from-layer="241" from-port="0" to-layer="242" to-port="3" />
-		<edge from-layer="242" from-port="4" to-layer="244" to-port="2" />
-		<edge from-layer="243" from-port="0" to-layer="244" to-port="3" />
-		<edge from-layer="244" from-port="4" to-layer="248" to-port="0" />
-		<edge from-layer="245" from-port="0" to-layer="246" to-port="0" />
-		<edge from-layer="246" from-port="1" to-layer="248" to-port="1" />
-		<edge from-layer="247" from-port="0" to-layer="248" to-port="2" />
-		<edge from-layer="248" from-port="3" to-layer="267" to-port="0" />
-		<edge from-layer="249" from-port="0" to-layer="250" to-port="0" />
-		<edge from-layer="250" from-port="1" to-layer="264" to-port="0" />
-		<edge from-layer="251" from-port="1" to-layer="254" to-port="0" />
-		<edge from-layer="252" from-port="0" to-layer="254" to-port="1" />
-		<edge from-layer="253" from-port="0" to-layer="254" to-port="2" />
-		<edge from-layer="254" from-port="3" to-layer="256" to-port="0" />
-		<edge from-layer="255" from-port="0" to-layer="2462" to-port="1" />
-		<edge from-layer="255" from-port="0" to-layer="1240" to-port="3" />
-		<edge from-layer="255" from-port="0" to-layer="3672" to-port="3" />
-		<edge from-layer="255" from-port="0" to-layer="1724" to-port="3" />
-		<edge from-layer="255" from-port="0" to-layer="7828" to-port="3" />
-		<edge from-layer="255" from-port="0" to-layer="7828" to-port="2" />
-		<edge from-layer="255" from-port="0" to-layer="7828" to-port="1" />
-		<edge from-layer="255" from-port="0" to-layer="2946" to-port="3" />
-		<edge from-layer="255" from-port="0" to-layer="2946" to-port="2" />
-		<edge from-layer="255" from-port="0" to-layer="2208" to-port="3" />
-		<edge from-layer="255" from-port="0" to-layer="2208" to-port="2" />
-		<edge from-layer="255" from-port="0" to-layer="7544" to-port="3" />
-		<edge from-layer="255" from-port="0" to-layer="7544" to-port="2" />
-		<edge from-layer="255" from-port="0" to-layer="2946" to-port="1" />
-		<edge from-layer="255" from-port="0" to-layer="3430" to-port="2" />
-		<edge from-layer="255" from-port="0" to-layer="256" to-port="1" />
-		<edge from-layer="255" from-port="0" to-layer="756" to-port="3" />
-		<edge from-layer="255" from-port="0" to-layer="756" to-port="2" />
-		<edge from-layer="255" from-port="0" to-layer="756" to-port="1" />
-		<edge from-layer="255" from-port="0" to-layer="3188" to-port="1" />
-		<edge from-layer="255" from-port="0" to-layer="3188" to-port="2" />
-		<edge from-layer="255" from-port="0" to-layer="3188" to-port="3" />
-		<edge from-layer="255" from-port="0" to-layer="4882" to-port="3" />
-		<edge from-layer="255" from-port="0" to-layer="3430" to-port="1" />
-		<edge from-layer="255" from-port="0" to-layer="1482" to-port="1" />
-		<edge from-layer="255" from-port="0" to-layer="1482" to-port="2" />
-		<edge from-layer="255" from-port="0" to-layer="1482" to-port="3" />
-		<edge from-layer="255" from-port="0" to-layer="1724" to-port="2" />
-		<edge from-layer="255" from-port="0" to-layer="3430" to-port="3" />
-		<edge from-layer="255" from-port="0" to-layer="4882" to-port="2" />
-		<edge from-layer="255" from-port="0" to-layer="4882" to-port="1" />
-		<edge from-layer="255" from-port="0" to-layer="4640" to-port="2" />
-		<edge from-layer="255" from-port="0" to-layer="3672" to-port="1" />
-		<edge from-layer="255" from-port="0" to-layer="3672" to-port="2" />
-		<edge from-layer="255" from-port="0" to-layer="1240" to-port="2" />
-		<edge from-layer="255" from-port="0" to-layer="2462" to-port="3" />
-		<edge from-layer="255" from-port="0" to-layer="2462" to-port="2" />
-		<edge from-layer="255" from-port="0" to-layer="1240" to-port="1" />
-		<edge from-layer="255" from-port="0" to-layer="1724" to-port="1" />
-		<edge from-layer="255" from-port="0" to-layer="514" to-port="1" />
-		<edge from-layer="255" from-port="0" to-layer="6576" to-port="1" />
-		<edge from-layer="255" from-port="0" to-layer="272" to-port="3" />
-		<edge from-layer="255" from-port="0" to-layer="6334" to-port="3" />
-		<edge from-layer="255" from-port="0" to-layer="6334" to-port="2" />
-		<edge from-layer="255" from-port="0" to-layer="6334" to-port="1" />
-		<edge from-layer="255" from-port="0" to-layer="998" to-port="3" />
-		<edge from-layer="255" from-port="0" to-layer="5366" to-port="1" />
-		<edge from-layer="255" from-port="0" to-layer="5366" to-port="2" />
-		<edge from-layer="255" from-port="0" to-layer="998" to-port="2" />
-		<edge from-layer="255" from-port="0" to-layer="6092" to-port="3" />
-		<edge from-layer="255" from-port="0" to-layer="6092" to-port="2" />
-		<edge from-layer="255" from-port="0" to-layer="998" to-port="1" />
-		<edge from-layer="255" from-port="0" to-layer="4398" to-port="1" />
-		<edge from-layer="255" from-port="0" to-layer="4398" to-port="2" />
-		<edge from-layer="255" from-port="0" to-layer="1966" to-port="1" />
-		<edge from-layer="255" from-port="0" to-layer="4398" to-port="3" />
-		<edge from-layer="255" from-port="0" to-layer="514" to-port="2" />
-		<edge from-layer="255" from-port="0" to-layer="514" to-port="3" />
-		<edge from-layer="255" from-port="0" to-layer="6092" to-port="1" />
-		<edge from-layer="255" from-port="0" to-layer="2704" to-port="1" />
-		<edge from-layer="255" from-port="0" to-layer="5366" to-port="3" />
-		<edge from-layer="255" from-port="0" to-layer="5850" to-port="3" />
-		<edge from-layer="255" from-port="0" to-layer="5850" to-port="2" />
-		<edge from-layer="255" from-port="0" to-layer="5850" to-port="1" />
-		<edge from-layer="255" from-port="0" to-layer="5608" to-port="3" />
-		<edge from-layer="255" from-port="0" to-layer="5608" to-port="2" />
-		<edge from-layer="255" from-port="0" to-layer="5608" to-port="1" />
-		<edge from-layer="255" from-port="0" to-layer="4640" to-port="1" />
-		<edge from-layer="255" from-port="0" to-layer="4640" to-port="3" />
-		<edge from-layer="255" from-port="0" to-layer="6576" to-port="3" />
-		<edge from-layer="255" from-port="0" to-layer="7544" to-port="1" />
-		<edge from-layer="255" from-port="0" to-layer="1966" to-port="2" />
-		<edge from-layer="255" from-port="0" to-layer="1966" to-port="3" />
-		<edge from-layer="255" from-port="0" to-layer="2208" to-port="1" />
-		<edge from-layer="255" from-port="0" to-layer="3914" to-port="3" />
-		<edge from-layer="255" from-port="0" to-layer="3914" to-port="2" />
-		<edge from-layer="255" from-port="0" to-layer="3914" to-port="1" />
-		<edge from-layer="255" from-port="0" to-layer="7302" to-port="3" />
-		<edge from-layer="255" from-port="0" to-layer="7302" to-port="2" />
-		<edge from-layer="255" from-port="0" to-layer="7302" to-port="1" />
-		<edge from-layer="255" from-port="0" to-layer="2704" to-port="3" />
-		<edge from-layer="255" from-port="0" to-layer="272" to-port="1" />
-		<edge from-layer="255" from-port="0" to-layer="5124" to-port="1" />
-		<edge from-layer="255" from-port="0" to-layer="7060" to-port="3" />
-		<edge from-layer="255" from-port="0" to-layer="7060" to-port="2" />
-		<edge from-layer="255" from-port="0" to-layer="7060" to-port="1" />
-		<edge from-layer="255" from-port="0" to-layer="2704" to-port="2" />
-		<edge from-layer="255" from-port="0" to-layer="272" to-port="2" />
-		<edge from-layer="255" from-port="0" to-layer="5124" to-port="2" />
-		<edge from-layer="255" from-port="0" to-layer="6818" to-port="3" />
-		<edge from-layer="255" from-port="0" to-layer="6818" to-port="2" />
-		<edge from-layer="255" from-port="0" to-layer="6818" to-port="1" />
-		<edge from-layer="255" from-port="0" to-layer="5124" to-port="3" />
-		<edge from-layer="255" from-port="0" to-layer="4156" to-port="1" />
-		<edge from-layer="255" from-port="0" to-layer="4156" to-port="2" />
-		<edge from-layer="255" from-port="0" to-layer="4156" to-port="3" />
-		<edge from-layer="255" from-port="0" to-layer="6576" to-port="2" />
-		<edge from-layer="256" from-port="2" to-layer="257" to-port="1" />
-		<edge from-layer="257" from-port="2" to-layer="258" to-port="0" />
-		<edge from-layer="258" from-port="2" to-layer="259" to-port="0" />
+		<edge from-layer="144" from-port="5" to-layer="145" to-port="0" />
+		<edge from-layer="144" from-port="5" to-layer="149" to-port="0" />
+		<edge from-layer="144" from-port="5" to-layer="155" to-port="0" />
+		<edge from-layer="145" from-port="2" to-layer="158" to-port="0" />
+		<edge from-layer="146" from-port="0" to-layer="149" to-port="1" />
+		<edge from-layer="147" from-port="0" to-layer="149" to-port="2" />
+		<edge from-layer="148" from-port="0" to-layer="149" to-port="3" />
+		<edge from-layer="149" from-port="5" to-layer="151" to-port="0" />
+		<edge from-layer="150" from-port="0" to-layer="151" to-port="1" />
+		<edge from-layer="151" from-port="2" to-layer="156" to-port="0" />
+		<edge from-layer="152" from-port="0" to-layer="155" to-port="1" />
+		<edge from-layer="153" from-port="0" to-layer="155" to-port="2" />
+		<edge from-layer="154" from-port="0" to-layer="155" to-port="3" />
+		<edge from-layer="155" from-port="5" to-layer="156" to-port="1" />
+		<edge from-layer="156" from-port="2" to-layer="157" to-port="0" />
+		<edge from-layer="157" from-port="2" to-layer="158" to-port="1" />
+		<edge from-layer="158" from-port="2" to-layer="163" to-port="0" />
+		<edge from-layer="159" from-port="0" to-layer="162" to-port="1" />
+		<edge from-layer="160" from-port="0" to-layer="162" to-port="2" />
+		<edge from-layer="161" from-port="0" to-layer="162" to-port="3" />
+		<edge from-layer="162" from-port="5" to-layer="163" to-port="1" />
+		<edge from-layer="163" from-port="2" to-layer="164" to-port="1" />
+		<edge from-layer="164" from-port="2" to-layer="273" to-port="1" />
+		<edge from-layer="164" from-port="2" to-layer="5159" to-port="0" />
+		<edge from-layer="165" from-port="0" to-layer="170" to-port="0" />
+		<edge from-layer="166" from-port="0" to-layer="169" to-port="1" />
+		<edge from-layer="167" from-port="0" to-layer="169" to-port="2" />
+		<edge from-layer="168" from-port="0" to-layer="169" to-port="3" />
+		<edge from-layer="169" from-port="4" to-layer="170" to-port="1" />
+		<edge from-layer="170" from-port="2" to-layer="171" to-port="0" />
+		<edge from-layer="171" from-port="1" to-layer="173" to-port="0" />
+		<edge from-layer="172" from-port="0" to-layer="173" to-port="2" />
+		<edge from-layer="173" from-port="3" to-layer="180" to-port="0" />
+		<edge from-layer="174" from-port="0" to-layer="177" to-port="1" />
+		<edge from-layer="175" from-port="0" to-layer="177" to-port="2" />
+		<edge from-layer="176" from-port="0" to-layer="177" to-port="3" />
+		<edge from-layer="177" from-port="5" to-layer="179" to-port="0" />
+		<edge from-layer="178" from-port="0" to-layer="179" to-port="1" />
+		<edge from-layer="179" from-port="2" to-layer="180" to-port="1" />
+		<edge from-layer="180" from-port="2" to-layer="273" to-port="2" />
+		<edge from-layer="180" from-port="2" to-layer="5158" to-port="0" />
+		<edge from-layer="181" from-port="0" to-layer="188" to-port="0" />
+		<edge from-layer="181" from-port="0" to-layer="189" to-port="1" />
+		<edge from-layer="182" from-port="1" to-layer="185" to-port="0" />
+		<edge from-layer="183" from-port="0" to-layer="185" to-port="1" />
+		<edge from-layer="184" from-port="0" to-layer="185" to-port="2" />
+		<edge from-layer="185" from-port="3" to-layer="211" to-port="1" />
+		<edge from-layer="185" from-port="3" to-layer="186" to-port="0" />
+		<edge from-layer="185" from-port="3" to-layer="204" to-port="0" />
+		<edge from-layer="186" from-port="1" to-layer="188" to-port="1" />
+		<edge from-layer="187" from-port="0" to-layer="197" to-port="2" />
+		<edge from-layer="187" from-port="0" to-layer="198" to-port="1" />
+		<edge from-layer="187" from-port="0" to-layer="188" to-port="2" />
+		<edge from-layer="188" from-port="3" to-layer="189" to-port="0" />
+		<edge from-layer="189" from-port="2" to-layer="199" to-port="0" />
+		<edge from-layer="190" from-port="0" to-layer="196" to-port="1" />
+		<edge from-layer="190" from-port="0" to-layer="197" to-port="0" />
+		<edge from-layer="191" from-port="1" to-layer="194" to-port="0" />
+		<edge from-layer="191" from-port="1" to-layer="231" to-port="0" />
+		<edge from-layer="192" from-port="0" to-layer="194" to-port="1" />
+		<edge from-layer="193" from-port="0" to-layer="194" to-port="2" />
+		<edge from-layer="194" from-port="3" to-layer="195" to-port="0" />
+		<edge from-layer="194" from-port="3" to-layer="202" to-port="0" />
+		<edge from-layer="194" from-port="3" to-layer="218" to-port="1" />
+		<edge from-layer="195" from-port="1" to-layer="196" to-port="0" />
+		<edge from-layer="196" from-port="2" to-layer="197" to-port="1" />
+		<edge from-layer="197" from-port="3" to-layer="198" to-port="0" />
+		<edge from-layer="198" from-port="2" to-layer="199" to-port="1" />
+		<edge from-layer="199" from-port="2" to-layer="208" to-port="0" />
+		<edge from-layer="200" from-port="0" to-layer="206" to-port="0" />
+		<edge from-layer="200" from-port="0" to-layer="260" to-port="1" />
+		<edge from-layer="201" from-port="0" to-layer="202" to-port="1" />
+		<edge from-layer="202" from-port="2" to-layer="4074" to-port="1" />
+		<edge from-layer="202" from-port="2" to-layer="1862" to-port="1" />
+		<edge from-layer="202" from-port="2" to-layer="1376" to-port="1" />
+		<edge from-layer="202" from-port="2" to-layer="4548" to-port="1" />
+		<edge from-layer="202" from-port="2" to-layer="1704" to-port="1" />
+		<edge from-layer="202" from-port="2" to-layer="2810" to-port="1" />
+		<edge from-layer="202" from-port="2" to-layer="1534" to-port="1" />
+		<edge from-layer="202" from-port="2" to-layer="3916" to-port="1" />
+		<edge from-layer="202" from-port="2" to-layer="3758" to-port="1" />
+		<edge from-layer="202" from-port="2" to-layer="4390" to-port="1" />
+		<edge from-layer="202" from-port="2" to-layer="270" to-port="1" />
+		<edge from-layer="202" from-port="2" to-layer="2020" to-port="1" />
+		<edge from-layer="202" from-port="2" to-layer="2178" to-port="1" />
+		<edge from-layer="202" from-port="2" to-layer="4706" to-port="1" />
+		<edge from-layer="202" from-port="2" to-layer="2336" to-port="1" />
+		<edge from-layer="202" from-port="2" to-layer="2494" to-port="1" />
+		<edge from-layer="202" from-port="2" to-layer="428" to-port="1" />
+		<edge from-layer="202" from-port="2" to-layer="5223" to-port="1" />
+		<edge from-layer="202" from-port="2" to-layer="902" to-port="1" />
+		<edge from-layer="202" from-port="2" to-layer="2652" to-port="1" />
+		<edge from-layer="202" from-port="2" to-layer="5022" to-port="1" />
+		<edge from-layer="202" from-port="2" to-layer="4232" to-port="1" />
+		<edge from-layer="202" from-port="2" to-layer="4864" to-port="1" />
+		<edge from-layer="202" from-port="2" to-layer="1060" to-port="1" />
+		<edge from-layer="202" from-port="2" to-layer="3600" to-port="1" />
+		<edge from-layer="202" from-port="2" to-layer="586" to-port="1" />
+		<edge from-layer="202" from-port="2" to-layer="744" to-port="1" />
+		<edge from-layer="202" from-port="2" to-layer="3126" to-port="1" />
+		<edge from-layer="202" from-port="2" to-layer="1218" to-port="1" />
+		<edge from-layer="202" from-port="2" to-layer="3284" to-port="1" />
+		<edge from-layer="202" from-port="2" to-layer="2968" to-port="1" />
+		<edge from-layer="202" from-port="2" to-layer="3442" to-port="1" />
+		<edge from-layer="202" from-port="2" to-layer="205" to-port="0" />
+		<edge from-layer="203" from-port="0" to-layer="204" to-port="1" />
+		<edge from-layer="204" from-port="2" to-layer="247" to-port="2" />
+		<edge from-layer="204" from-port="2" to-layer="252" to-port="2" />
+		<edge from-layer="204" from-port="2" to-layer="205" to-port="1" />
+		<edge from-layer="205" from-port="2" to-layer="206" to-port="1" />
+		<edge from-layer="206" from-port="2" to-layer="208" to-port="1" />
+		<edge from-layer="207" from-port="0" to-layer="208" to-port="2" />
+		<edge from-layer="208" from-port="3" to-layer="225" to-port="0" />
+		<edge from-layer="209" from-port="0" to-layer="211" to-port="0" />
+		<edge from-layer="210" from-port="0" to-layer="211" to-port="2" />
+		<edge from-layer="211" from-port="3" to-layer="212" to-port="0" />
+		<edge from-layer="212" from-port="1" to-layer="213" to-port="0" />
+		<edge from-layer="213" from-port="1" to-layer="223" to-port="0" />
+		<edge from-layer="214" from-port="1" to-layer="217" to-port="0" />
+		<edge from-layer="215" from-port="0" to-layer="217" to-port="1" />
+		<edge from-layer="216" from-port="0" to-layer="217" to-port="2" />
+		<edge from-layer="217" from-port="3" to-layer="218" to-port="0" />
+		<edge from-layer="217" from-port="3" to-layer="220" to-port="0" />
+		<edge from-layer="217" from-port="3" to-layer="269" to-port="0" />
+		<edge from-layer="218" from-port="2" to-layer="220" to-port="1" />
+		<edge from-layer="219" from-port="0" to-layer="220" to-port="2" />
+		<edge from-layer="220" from-port="3" to-layer="222" to-port="0" />
+		<edge from-layer="221" from-port="0" to-layer="222" to-port="1" />
+		<edge from-layer="222" from-port="2" to-layer="223" to-port="1" />
+		<edge from-layer="223" from-port="2" to-layer="224" to-port="0" />
+		<edge from-layer="224" from-port="1" to-layer="225" to-port="1" />
+		<edge from-layer="225" from-port="2" to-layer="227" to-port="0" />
+		<edge from-layer="226" from-port="0" to-layer="227" to-port="1" />
+		<edge from-layer="227" from-port="2" to-layer="228" to-port="0" />
+		<edge from-layer="228" from-port="2" to-layer="236" to-port="0" />
+		<edge from-layer="229" from-port="0" to-layer="231" to-port="1" />
+		<edge from-layer="230" from-port="0" to-layer="231" to-port="2" />
+		<edge from-layer="231" from-port="3" to-layer="235" to-port="0" />
+		<edge from-layer="232" from-port="0" to-layer="235" to-port="1" />
+		<edge from-layer="233" from-port="0" to-layer="235" to-port="2" />
+		<edge from-layer="234" from-port="0" to-layer="235" to-port="3" />
+		<edge from-layer="235" from-port="4" to-layer="236" to-port="1" />
+		<edge from-layer="236" from-port="2" to-layer="238" to-port="0" />
+		<edge from-layer="236" from-port="2" to-layer="252" to-port="0" />
+		<edge from-layer="236" from-port="2" to-layer="240" to-port="0" />
+		<edge from-layer="237" from-port="0" to-layer="238" to-port="1" />
+		<edge from-layer="238" from-port="2" to-layer="265" to-port="0" />
+		<edge from-layer="239" from-port="0" to-layer="241" to-port="1" />
+		<edge from-layer="239" from-port="0" to-layer="243" to-port="0" />
+		<edge from-layer="240" from-port="1" to-layer="241" to-port="0" />
+		<edge from-layer="240" from-port="1" to-layer="266" to-port="1" />
+		<edge from-layer="240" from-port="1" to-layer="244" to-port="1" />
+		<edge from-layer="241" from-port="2" to-layer="243" to-port="1" />
+		<edge from-layer="242" from-port="0" to-layer="243" to-port="2" />
+		<edge from-layer="243" from-port="3" to-layer="244" to-port="0" />
+		<edge from-layer="244" from-port="2" to-layer="247" to-port="0" />
+		<edge from-layer="245" from-port="0" to-layer="247" to-port="1" />
+		<edge from-layer="246" from-port="0" to-layer="247" to-port="3" />
+		<edge from-layer="247" from-port="5" to-layer="249" to-port="0" />
+		<edge from-layer="248" from-port="0" to-layer="249" to-port="1" />
+		<edge from-layer="249" from-port="2" to-layer="265" to-port="1" />
+		<edge from-layer="250" from-port="0" to-layer="252" to-port="1" />
+		<edge from-layer="251" from-port="0" to-layer="252" to-port="3" />
+		<edge from-layer="252" from-port="5" to-layer="261" to-port="0" />
+		<edge from-layer="252" from-port="5" to-layer="260" to-port="2" />
+		<edge from-layer="252" from-port="5" to-layer="257" to-port="0" />
+		<edge from-layer="253" from-port="2" to-layer="255" to-port="0" />
+		<edge from-layer="254" from-port="0" to-layer="255" to-port="1" />
+		<edge from-layer="255" from-port="2" to-layer="256" to-port="0" />
+		<edge from-layer="256" from-port="1" to-layer="257" to-port="1" />
+		<edge from-layer="257" from-port="2" to-layer="259" to-port="0" />
+		<edge from-layer="258" from-port="0" to-layer="259" to-port="1" />
 		<edge from-layer="259" from-port="2" to-layer="260" to-port="0" />
-		<edge from-layer="260" from-port="1" to-layer="263" to-port="0" />
-		<edge from-layer="261" from-port="0" to-layer="262" to-port="0" />
-		<edge from-layer="262" from-port="1" to-layer="263" to-port="1" />
-		<edge from-layer="263" from-port="2" to-layer="264" to-port="1" />
-		<edge from-layer="264" from-port="2" to-layer="266" to-port="0" />
-		<edge from-layer="265" from-port="0" to-layer="266" to-port="1" />
-		<edge from-layer="266" from-port="2" to-layer="5845" to-port="1" />
-		<edge from-layer="266" from-port="2" to-layer="6329" to-port="1" />
-		<edge from-layer="266" from-port="2" to-layer="4877" to-port="1" />
-		<edge from-layer="266" from-port="2" to-layer="509" to-port="1" />
-		<edge from-layer="266" from-port="2" to-layer="4393" to-port="1" />
-		<edge from-layer="266" from-port="2" to-layer="993" to-port="1" />
-		<edge from-layer="266" from-port="2" to-layer="6087" to-port="1" />
-		<edge from-layer="266" from-port="2" to-layer="2699" to-port="1" />
-		<edge from-layer="266" from-port="2" to-layer="5119" to-port="1" />
-		<edge from-layer="266" from-port="2" to-layer="5603" to-port="1" />
-		<edge from-layer="266" from-port="2" to-layer="4635" to-port="1" />
-		<edge from-layer="266" from-port="2" to-layer="4151" to-port="1" />
-		<edge from-layer="266" from-port="2" to-layer="5361" to-port="1" />
-		<edge from-layer="266" from-port="2" to-layer="3909" to-port="1" />
-		<edge from-layer="266" from-port="2" to-layer="2203" to-port="1" />
-		<edge from-layer="266" from-port="2" to-layer="2457" to-port="1" />
-		<edge from-layer="266" from-port="2" to-layer="751" to-port="1" />
-		<edge from-layer="266" from-port="2" to-layer="3425" to-port="1" />
-		<edge from-layer="266" from-port="2" to-layer="1477" to-port="1" />
-		<edge from-layer="266" from-port="2" to-layer="3667" to-port="1" />
-		<edge from-layer="266" from-port="2" to-layer="1719" to-port="1" />
-		<edge from-layer="266" from-port="2" to-layer="1235" to-port="1" />
-		<edge from-layer="266" from-port="2" to-layer="7823" to-port="1" />
-		<edge from-layer="266" from-port="2" to-layer="6571" to-port="1" />
-		<edge from-layer="266" from-port="2" to-layer="1961" to-port="1" />
-		<edge from-layer="266" from-port="2" to-layer="2941" to-port="1" />
-		<edge from-layer="266" from-port="2" to-layer="7539" to-port="1" />
-		<edge from-layer="266" from-port="2" to-layer="267" to-port="1" />
-		<edge from-layer="266" from-port="2" to-layer="7297" to-port="1" />
-		<edge from-layer="266" from-port="2" to-layer="7055" to-port="1" />
-		<edge from-layer="266" from-port="2" to-layer="6813" to-port="1" />
-		<edge from-layer="266" from-port="2" to-layer="3183" to-port="1" />
-		<edge from-layer="267" from-port="2" to-layer="275" to-port="0" />
-		<edge from-layer="267" from-port="2" to-layer="268" to-port="0" />
-		<edge from-layer="268" from-port="1" to-layer="274" to-port="0" />
-		<edge from-layer="269" from-port="0" to-layer="271" to-port="1" />
-		<edge from-layer="270" from-port="0" to-layer="271" to-port="2" />
-		<edge from-layer="271" from-port="3" to-layer="272" to-port="0" />
-		<edge from-layer="272" from-port="4" to-layer="273" to-port="0" />
-		<edge from-layer="273" from-port="1" to-layer="274" to-port="1" />
-		<edge from-layer="274" from-port="2" to-layer="275" to-port="1" />
-		<edge from-layer="275" from-port="2" to-layer="276" to-port="3" />
-		<edge from-layer="276" from-port="4" to-layer="278" to-port="0" />
-		<edge from-layer="277" from-port="0" to-layer="278" to-port="1" />
-		<edge from-layer="278" from-port="2" to-layer="280" to-port="0" />
-		<edge from-layer="279" from-port="0" to-layer="280" to-port="1" />
-		<edge from-layer="280" from-port="2" to-layer="289" to-port="0" />
-		<edge from-layer="281" from-port="0" to-layer="282" to-port="0" />
-		<edge from-layer="282" from-port="1" to-layer="285" to-port="0" />
-		<edge from-layer="283" from-port="0" to-layer="284" to-port="0" />
-		<edge from-layer="284" from-port="1" to-layer="285" to-port="1" />
-		<edge from-layer="285" from-port="2" to-layer="287" to-port="0" />
-		<edge from-layer="286" from-port="0" to-layer="287" to-port="1" />
-		<edge from-layer="287" from-port="2" to-layer="288" to-port="0" />
-		<edge from-layer="288" from-port="1" to-layer="289" to-port="1" />
-		<edge from-layer="289" from-port="2" to-layer="292" to-port="0" />
-		<edge from-layer="290" from-port="0" to-layer="291" to-port="0" />
-		<edge from-layer="291" from-port="1" to-layer="292" to-port="1" />
-		<edge from-layer="292" from-port="2" to-layer="293" to-port="1" />
-		<edge from-layer="293" from-port="2" to-layer="294" to-port="0" />
+		<edge from-layer="260" from-port="3" to-layer="262" to-port="0" />
+		<edge from-layer="261" from-port="1" to-layer="262" to-port="1" />
+		<edge from-layer="262" from-port="2" to-layer="264" to-port="0" />
+		<edge from-layer="263" from-port="0" to-layer="264" to-port="1" />
+		<edge from-layer="264" from-port="2" to-layer="265" to-port="2" />
+		<edge from-layer="265" from-port="3" to-layer="266" to-port="0" />
+		<edge from-layer="266" from-port="2" to-layer="2654" to-port="0" />
+		<edge from-layer="266" from-port="2" to-layer="588" to-port="0" />
+		<edge from-layer="266" from-port="2" to-layer="4392" to-port="0" />
+		<edge from-layer="266" from-port="2" to-layer="1378" to-port="0" />
+		<edge from-layer="266" from-port="2" to-layer="4550" to-port="0" />
+		<edge from-layer="266" from-port="2" to-layer="1536" to-port="0" />
+		<edge from-layer="266" from-port="2" to-layer="4708" to-port="0" />
+		<edge from-layer="266" from-port="2" to-layer="904" to-port="0" />
+		<edge from-layer="266" from-port="2" to-layer="4866" to-port="0" />
+		<edge from-layer="266" from-port="2" to-layer="1220" to-port="0" />
+		<edge from-layer="266" from-port="2" to-layer="5024" to-port="0" />
+		<edge from-layer="266" from-port="2" to-layer="2022" to-port="0" />
+		<edge from-layer="266" from-port="2" to-layer="1062" to-port="0" />
+		<edge from-layer="266" from-port="2" to-layer="5225" to-port="0" />
+		<edge from-layer="266" from-port="2" to-layer="430" to-port="0" />
+		<edge from-layer="266" from-port="2" to-layer="2496" to-port="0" />
+		<edge from-layer="266" from-port="2" to-layer="1706" to-port="0" />
+		<edge from-layer="266" from-port="2" to-layer="2338" to-port="0" />
+		<edge from-layer="266" from-port="2" to-layer="1864" to-port="0" />
+		<edge from-layer="266" from-port="2" to-layer="2180" to-port="0" />
+		<edge from-layer="266" from-port="2" to-layer="2970" to-port="0" />
+		<edge from-layer="266" from-port="2" to-layer="272" to-port="0" />
+		<edge from-layer="266" from-port="2" to-layer="4076" to-port="0" />
+		<edge from-layer="266" from-port="2" to-layer="2812" to-port="0" />
+		<edge from-layer="266" from-port="2" to-layer="3918" to-port="0" />
+		<edge from-layer="266" from-port="2" to-layer="3760" to-port="0" />
+		<edge from-layer="266" from-port="2" to-layer="746" to-port="0" />
+		<edge from-layer="266" from-port="2" to-layer="3602" to-port="0" />
+		<edge from-layer="266" from-port="2" to-layer="3444" to-port="0" />
+		<edge from-layer="266" from-port="2" to-layer="4234" to-port="0" />
+		<edge from-layer="266" from-port="2" to-layer="3286" to-port="0" />
+		<edge from-layer="266" from-port="2" to-layer="3128" to-port="0" />
+		<edge from-layer="267" from-port="0" to-layer="272" to-port="1" />
+		<edge from-layer="268" from-port="0" to-layer="269" to-port="1" />
+		<edge from-layer="269" from-port="2" to-layer="270" to-port="0" />
+		<edge from-layer="270" from-port="2" to-layer="272" to-port="2" />
+		<edge from-layer="271" from-port="0" to-layer="272" to-port="3" />
+		<edge from-layer="272" from-port="5" to-layer="273" to-port="3" />
+		<edge from-layer="273" from-port="4" to-layer="275" to-port="0" />
+		<edge from-layer="274" from-port="0" to-layer="275" to-port="1" />
+		<edge from-layer="275" from-port="2" to-layer="277" to-port="0" />
+		<edge from-layer="276" from-port="0" to-layer="277" to-port="1" />
+		<edge from-layer="277" from-port="2" to-layer="286" to-port="0" />
+		<edge from-layer="278" from-port="0" to-layer="279" to-port="0" />
+		<edge from-layer="279" from-port="1" to-layer="282" to-port="0" />
+		<edge from-layer="280" from-port="0" to-layer="281" to-port="0" />
+		<edge from-layer="281" from-port="1" to-layer="282" to-port="1" />
+		<edge from-layer="282" from-port="2" to-layer="284" to-port="0" />
+		<edge from-layer="283" from-port="0" to-layer="284" to-port="1" />
+		<edge from-layer="284" from-port="2" to-layer="285" to-port="0" />
+		<edge from-layer="285" from-port="1" to-layer="286" to-port="1" />
+		<edge from-layer="286" from-port="2" to-layer="288" to-port="0" />
+		<edge from-layer="287" from-port="0" to-layer="288" to-port="1" />
+		<edge from-layer="288" from-port="2" to-layer="289" to-port="1" />
+		<edge from-layer="289" from-port="2" to-layer="290" to-port="0" />
+		<edge from-layer="290" from-port="2" to-layer="292" to-port="0" />
+		<edge from-layer="290" from-port="2" to-layer="448" to-port="1" />
+		<edge from-layer="291" from-port="0" to-layer="292" to-port="1" />
+		<edge from-layer="292" from-port="2" to-layer="323" to-port="0" />
+		<edge from-layer="292" from-port="2" to-layer="294" to-port="0" />
+		<edge from-layer="293" from-port="0" to-layer="294" to-port="1" />
 		<edge from-layer="294" from-port="2" to-layer="296" to-port="0" />
-		<edge from-layer="294" from-port="2" to-layer="536" to-port="1" />
 		<edge from-layer="295" from-port="0" to-layer="296" to-port="1" />
-		<edge from-layer="296" from-port="2" to-layer="299" to-port="0" />
-		<edge from-layer="296" from-port="2" to-layer="330" to-port="0" />
+		<edge from-layer="296" from-port="2" to-layer="305" to-port="0" />
 		<edge from-layer="297" from-port="0" to-layer="298" to-port="0" />
-		<edge from-layer="298" from-port="1" to-layer="299" to-port="1" />
-		<edge from-layer="299" from-port="2" to-layer="302" to-port="0" />
-		<edge from-layer="300" from-port="0" to-layer="301" to-port="0" />
-		<edge from-layer="301" from-port="1" to-layer="302" to-port="1" />
-		<edge from-layer="302" from-port="2" to-layer="311" to-port="0" />
-		<edge from-layer="303" from-port="0" to-layer="304" to-port="0" />
-		<edge from-layer="304" from-port="1" to-layer="307" to-port="0" />
-		<edge from-layer="305" from-port="0" to-layer="306" to-port="0" />
-		<edge from-layer="306" from-port="1" to-layer="307" to-port="1" />
-		<edge from-layer="307" from-port="2" to-layer="309" to-port="0" />
-		<edge from-layer="308" from-port="0" to-layer="309" to-port="1" />
-		<edge from-layer="309" from-port="2" to-layer="310" to-port="0" />
-		<edge from-layer="310" from-port="1" to-layer="311" to-port="1" />
-		<edge from-layer="311" from-port="2" to-layer="314" to-port="0" />
+		<edge from-layer="298" from-port="1" to-layer="301" to-port="0" />
+		<edge from-layer="299" from-port="0" to-layer="300" to-port="0" />
+		<edge from-layer="300" from-port="1" to-layer="301" to-port="1" />
+		<edge from-layer="301" from-port="2" to-layer="303" to-port="0" />
+		<edge from-layer="302" from-port="0" to-layer="303" to-port="1" />
+		<edge from-layer="303" from-port="2" to-layer="304" to-port="0" />
+		<edge from-layer="304" from-port="1" to-layer="305" to-port="1" />
+		<edge from-layer="305" from-port="2" to-layer="308" to-port="0" />
+		<edge from-layer="306" from-port="0" to-layer="307" to-port="0" />
+		<edge from-layer="307" from-port="1" to-layer="308" to-port="1" />
+		<edge from-layer="308" from-port="2" to-layer="309" to-port="0" />
+		<edge from-layer="309" from-port="1" to-layer="318" to-port="0" />
+		<edge from-layer="310" from-port="0" to-layer="311" to-port="0" />
+		<edge from-layer="311" from-port="1" to-layer="314" to-port="0" />
 		<edge from-layer="312" from-port="0" to-layer="313" to-port="0" />
 		<edge from-layer="313" from-port="1" to-layer="314" to-port="1" />
-		<edge from-layer="314" from-port="2" to-layer="315" to-port="0" />
-		<edge from-layer="315" from-port="1" to-layer="324" to-port="0" />
-		<edge from-layer="316" from-port="0" to-layer="317" to-port="0" />
-		<edge from-layer="317" from-port="1" to-layer="320" to-port="0" />
-		<edge from-layer="318" from-port="0" to-layer="319" to-port="0" />
-		<edge from-layer="319" from-port="1" to-layer="320" to-port="1" />
-		<edge from-layer="320" from-port="2" to-layer="322" to-port="0" />
-		<edge from-layer="321" from-port="0" to-layer="322" to-port="1" />
-		<edge from-layer="322" from-port="2" to-layer="323" to-port="0" />
-		<edge from-layer="323" from-port="1" to-layer="324" to-port="1" />
-		<edge from-layer="324" from-port="2" to-layer="327" to-port="0" />
-		<edge from-layer="325" from-port="0" to-layer="326" to-port="0" />
-		<edge from-layer="326" from-port="1" to-layer="327" to-port="1" />
-		<edge from-layer="327" from-port="2" to-layer="535" to-port="0" />
+		<edge from-layer="314" from-port="2" to-layer="316" to-port="0" />
+		<edge from-layer="315" from-port="0" to-layer="316" to-port="1" />
+		<edge from-layer="316" from-port="2" to-layer="317" to-port="0" />
+		<edge from-layer="317" from-port="1" to-layer="318" to-port="1" />
+		<edge from-layer="318" from-port="2" to-layer="321" to-port="0" />
+		<edge from-layer="319" from-port="0" to-layer="320" to-port="0" />
+		<edge from-layer="320" from-port="1" to-layer="321" to-port="1" />
+		<edge from-layer="321" from-port="2" to-layer="447" to-port="0" />
+		<edge from-layer="322" from-port="0" to-layer="323" to-port="1" />
+		<edge from-layer="323" from-port="2" to-layer="325" to-port="0" />
+		<edge from-layer="324" from-port="0" to-layer="325" to-port="1" />
+		<edge from-layer="325" from-port="2" to-layer="334" to-port="0" />
+		<edge from-layer="326" from-port="0" to-layer="327" to-port="0" />
+		<edge from-layer="327" from-port="1" to-layer="330" to-port="0" />
 		<edge from-layer="328" from-port="0" to-layer="329" to-port="0" />
 		<edge from-layer="329" from-port="1" to-layer="330" to-port="1" />
-		<edge from-layer="330" from-port="2" to-layer="333" to-port="0" />
-		<edge from-layer="331" from-port="0" to-layer="332" to-port="0" />
-		<edge from-layer="332" from-port="1" to-layer="333" to-port="1" />
-		<edge from-layer="333" from-port="2" to-layer="342" to-port="0" />
-		<edge from-layer="334" from-port="0" to-layer="335" to-port="0" />
-		<edge from-layer="335" from-port="1" to-layer="338" to-port="0" />
-		<edge from-layer="336" from-port="0" to-layer="337" to-port="0" />
-		<edge from-layer="337" from-port="1" to-layer="338" to-port="1" />
-		<edge from-layer="338" from-port="2" to-layer="340" to-port="0" />
-		<edge from-layer="339" from-port="0" to-layer="340" to-port="1" />
-		<edge from-layer="340" from-port="2" to-layer="341" to-port="0" />
-		<edge from-layer="341" from-port="1" to-layer="342" to-port="1" />
-		<edge from-layer="342" from-port="2" to-layer="345" to-port="0" />
-		<edge from-layer="343" from-port="0" to-layer="344" to-port="0" />
-		<edge from-layer="344" from-port="1" to-layer="345" to-port="1" />
-		<edge from-layer="345" from-port="2" to-layer="347" to-port="0" />
-		<edge from-layer="346" from-port="0" to-layer="347" to-port="1" />
-		<edge from-layer="347" from-port="2" to-layer="475" to-port="0" />
-		<edge from-layer="347" from-port="2" to-layer="351" to-port="0" />
-		<edge from-layer="347" from-port="2" to-layer="364" to-port="0" />
-		<edge from-layer="348" from-port="0" to-layer="351" to-port="1" />
-		<edge from-layer="349" from-port="0" to-layer="351" to-port="2" />
-		<edge from-layer="350" from-port="0" to-layer="351" to-port="3" />
-		<edge from-layer="351" from-port="4" to-layer="353" to-port="0" />
-		<edge from-layer="352" from-port="0" to-layer="353" to-port="1" />
-		<edge from-layer="353" from-port="2" to-layer="426" to-port="0" />
-		<edge from-layer="353" from-port="2" to-layer="357" to-port="0" />
-		<edge from-layer="354" from-port="0" to-layer="357" to-port="1" />
-		<edge from-layer="355" from-port="0" to-layer="357" to-port="2" />
-		<edge from-layer="356" from-port="0" to-layer="357" to-port="3" />
-		<edge from-layer="357" from-port="4" to-layer="391" to-port="0" />
-		<edge from-layer="357" from-port="4" to-layer="394" to-port="0" />
-		<edge from-layer="357" from-port="4" to-layer="405" to-port="0" />
-		<edge from-layer="357" from-port="4" to-layer="411" to-port="0" />
-		<edge from-layer="358" from-port="0" to-layer="359" to-port="0" />
-		<edge from-layer="359" from-port="1" to-layer="387" to-port="0" />
-		<edge from-layer="360" from-port="0" to-layer="387" to-port="1" />
-		<edge from-layer="361" from-port="0" to-layer="364" to-port="1" />
-		<edge from-layer="362" from-port="0" to-layer="364" to-port="2" />
-		<edge from-layer="363" from-port="0" to-layer="364" to-port="3" />
-		<edge from-layer="364" from-port="4" to-layer="366" to-port="0" />
-		<edge from-layer="365" from-port="0" to-layer="366" to-port="1" />
-		<edge from-layer="366" from-port="2" to-layer="459" to-port="0" />
-		<edge from-layer="366" from-port="2" to-layer="431" to-port="0" />
-		<edge from-layer="366" from-port="2" to-layer="367" to-port="0" />
-		<edge from-layer="367" from-port="1" to-layer="369" to-port="0" />
-		<edge from-layer="368" from-port="0" to-layer="369" to-port="2" />
-		<edge from-layer="369" from-port="3" to-layer="383" to-port="0" />
-		<edge from-layer="370" from-port="0" to-layer="371" to-port="0" />
-		<edge from-layer="371" from-port="1" to-layer="376" to-port="0" />
-		<edge from-layer="372" from-port="0" to-layer="375" to-port="1" />
-		<edge from-layer="373" from-port="0" to-layer="375" to-port="2" />
-		<edge from-layer="374" from-port="0" to-layer="375" to-port="3" />
-		<edge from-layer="375" from-port="4" to-layer="376" to-port="1" />
-		<edge from-layer="376" from-port="2" to-layer="377" to-port="0" />
-		<edge from-layer="377" from-port="1" to-layer="379" to-port="0" />
-		<edge from-layer="378" from-port="0" to-layer="379" to-port="2" />
-		<edge from-layer="379" from-port="3" to-layer="380" to-port="0" />
-		<edge from-layer="379" from-port="3" to-layer="461" to-port="0" />
-		<edge from-layer="380" from-port="1" to-layer="382" to-port="0" />
-		<edge from-layer="381" from-port="0" to-layer="382" to-port="2" />
-		<edge from-layer="382" from-port="3" to-layer="383" to-port="1" />
-		<edge from-layer="383" from-port="2" to-layer="385" to-port="0" />
-		<edge from-layer="384" from-port="0" to-layer="385" to-port="1" />
-		<edge from-layer="385" from-port="2" to-layer="417" to-port="2" />
-		<edge from-layer="385" from-port="2" to-layer="387" to-port="2" />
-		<edge from-layer="386" from-port="0" to-layer="387" to-port="3" />
-		<edge from-layer="387" from-port="4" to-layer="389" to-port="0" />
-		<edge from-layer="388" from-port="0" to-layer="389" to-port="2" />
-		<edge from-layer="389" from-port="3" to-layer="390" to-port="0" />
-		<edge from-layer="390" from-port="2" to-layer="432" to-port="1" />
-		<edge from-layer="390" from-port="2" to-layer="391" to-port="1" />
-		<edge from-layer="391" from-port="2" to-layer="422" to-port="0" />
-		<edge from-layer="392" from-port="0" to-layer="402" to-port="0" />
-		<edge from-layer="393" from-port="0" to-layer="402" to-port="1" />
-		<edge from-layer="394" from-port="1" to-layer="396" to-port="0" />
-		<edge from-layer="395" from-port="0" to-layer="396" to-port="2" />
-		<edge from-layer="396" from-port="3" to-layer="397" to-port="0" />
-		<edge from-layer="397" from-port="2" to-layer="398" to-port="0" />
-		<edge from-layer="398" from-port="1" to-layer="400" to-port="0" />
-		<edge from-layer="399" from-port="0" to-layer="400" to-port="1" />
-		<edge from-layer="400" from-port="2" to-layer="402" to-port="2" />
-		<edge from-layer="401" from-port="0" to-layer="402" to-port="3" />
-		<edge from-layer="402" from-port="4" to-layer="411" to-port="2" />
-		<edge from-layer="402" from-port="4" to-layer="405" to-port="1" />
-		<edge from-layer="403" from-port="0" to-layer="405" to-port="2" />
-		<edge from-layer="404" from-port="0" to-layer="405" to-port="3" />
-		<edge from-layer="405" from-port="4" to-layer="408" to-port="0" />
-		<edge from-layer="406" from-port="0" to-layer="407" to-port="0" />
-		<edge from-layer="407" from-port="1" to-layer="408" to-port="1" />
-		<edge from-layer="408" from-port="2" to-layer="412" to-port="0" />
-		<edge from-layer="409" from-port="0" to-layer="411" to-port="1" />
+		<edge from-layer="330" from-port="2" to-layer="332" to-port="0" />
+		<edge from-layer="331" from-port="0" to-layer="332" to-port="1" />
+		<edge from-layer="332" from-port="2" to-layer="333" to-port="0" />
+		<edge from-layer="333" from-port="1" to-layer="334" to-port="1" />
+		<edge from-layer="334" from-port="2" to-layer="336" to-port="0" />
+		<edge from-layer="335" from-port="0" to-layer="336" to-port="1" />
+		<edge from-layer="336" from-port="2" to-layer="338" to-port="0" />
+		<edge from-layer="337" from-port="0" to-layer="338" to-port="1" />
+		<edge from-layer="338" from-port="2" to-layer="380" to-port="0" />
+		<edge from-layer="338" from-port="2" to-layer="419" to-port="0" />
+		<edge from-layer="338" from-port="2" to-layer="342" to-port="0" />
+		<edge from-layer="339" from-port="0" to-layer="342" to-port="1" />
+		<edge from-layer="340" from-port="0" to-layer="342" to-port="2" />
+		<edge from-layer="341" from-port="0" to-layer="342" to-port="3" />
+		<edge from-layer="342" from-port="5" to-layer="344" to-port="0" />
+		<edge from-layer="343" from-port="0" to-layer="344" to-port="1" />
+		<edge from-layer="344" from-port="2" to-layer="366" to-port="0" />
+		<edge from-layer="344" from-port="2" to-layer="348" to-port="0" />
+		<edge from-layer="345" from-port="0" to-layer="348" to-port="1" />
+		<edge from-layer="346" from-port="0" to-layer="348" to-port="2" />
+		<edge from-layer="347" from-port="0" to-layer="348" to-port="3" />
+		<edge from-layer="348" from-port="5" to-layer="359" to-port="0" />
+		<edge from-layer="348" from-port="5" to-layer="349" to-port="0" />
+		<edge from-layer="348" from-port="5" to-layer="353" to-port="0" />
+		<edge from-layer="349" from-port="2" to-layer="362" to-port="0" />
+		<edge from-layer="350" from-port="0" to-layer="353" to-port="1" />
+		<edge from-layer="351" from-port="0" to-layer="353" to-port="2" />
+		<edge from-layer="352" from-port="0" to-layer="353" to-port="3" />
+		<edge from-layer="353" from-port="5" to-layer="355" to-port="0" />
+		<edge from-layer="354" from-port="0" to-layer="355" to-port="1" />
+		<edge from-layer="355" from-port="2" to-layer="360" to-port="0" />
+		<edge from-layer="356" from-port="0" to-layer="359" to-port="1" />
+		<edge from-layer="357" from-port="0" to-layer="359" to-port="2" />
+		<edge from-layer="358" from-port="0" to-layer="359" to-port="3" />
+		<edge from-layer="359" from-port="5" to-layer="360" to-port="1" />
+		<edge from-layer="360" from-port="2" to-layer="361" to-port="0" />
+		<edge from-layer="361" from-port="2" to-layer="362" to-port="1" />
+		<edge from-layer="362" from-port="2" to-layer="367" to-port="0" />
+		<edge from-layer="363" from-port="0" to-layer="366" to-port="1" />
+		<edge from-layer="364" from-port="0" to-layer="366" to-port="2" />
+		<edge from-layer="365" from-port="0" to-layer="366" to-port="3" />
+		<edge from-layer="366" from-port="5" to-layer="367" to-port="1" />
+		<edge from-layer="367" from-port="2" to-layer="431" to-port="0" />
+		<edge from-layer="368" from-port="0" to-layer="373" to-port="0" />
+		<edge from-layer="369" from-port="0" to-layer="372" to-port="1" />
+		<edge from-layer="370" from-port="0" to-layer="372" to-port="2" />
+		<edge from-layer="371" from-port="0" to-layer="372" to-port="3" />
+		<edge from-layer="372" from-port="4" to-layer="373" to-port="1" />
+		<edge from-layer="373" from-port="2" to-layer="374" to-port="0" />
+		<edge from-layer="374" from-port="1" to-layer="376" to-port="0" />
+		<edge from-layer="375" from-port="0" to-layer="376" to-port="2" />
+		<edge from-layer="376" from-port="3" to-layer="406" to-port="0" />
+		<edge from-layer="376" from-port="3" to-layer="424" to-port="0" />
+		<edge from-layer="377" from-port="0" to-layer="380" to-port="1" />
+		<edge from-layer="378" from-port="0" to-layer="380" to-port="2" />
+		<edge from-layer="379" from-port="0" to-layer="380" to-port="3" />
+		<edge from-layer="380" from-port="5" to-layer="382" to-port="0" />
+		<edge from-layer="381" from-port="0" to-layer="382" to-port="1" />
+		<edge from-layer="382" from-port="2" to-layer="386" to-port="0" />
+		<edge from-layer="382" from-port="2" to-layer="404" to-port="0" />
+		<edge from-layer="383" from-port="0" to-layer="386" to-port="1" />
+		<edge from-layer="384" from-port="0" to-layer="386" to-port="2" />
+		<edge from-layer="385" from-port="0" to-layer="386" to-port="3" />
+		<edge from-layer="386" from-port="5" to-layer="397" to-port="0" />
+		<edge from-layer="386" from-port="5" to-layer="391" to-port="0" />
+		<edge from-layer="386" from-port="5" to-layer="387" to-port="0" />
+		<edge from-layer="387" from-port="2" to-layer="400" to-port="0" />
+		<edge from-layer="388" from-port="0" to-layer="391" to-port="1" />
+		<edge from-layer="389" from-port="0" to-layer="391" to-port="2" />
+		<edge from-layer="390" from-port="0" to-layer="391" to-port="3" />
+		<edge from-layer="391" from-port="5" to-layer="393" to-port="0" />
+		<edge from-layer="392" from-port="0" to-layer="393" to-port="1" />
+		<edge from-layer="393" from-port="2" to-layer="398" to-port="0" />
+		<edge from-layer="394" from-port="0" to-layer="397" to-port="1" />
+		<edge from-layer="395" from-port="0" to-layer="397" to-port="2" />
+		<edge from-layer="396" from-port="0" to-layer="397" to-port="3" />
+		<edge from-layer="397" from-port="5" to-layer="398" to-port="1" />
+		<edge from-layer="398" from-port="2" to-layer="399" to-port="0" />
+		<edge from-layer="399" from-port="2" to-layer="400" to-port="1" />
+		<edge from-layer="400" from-port="2" to-layer="405" to-port="0" />
+		<edge from-layer="401" from-port="0" to-layer="404" to-port="1" />
+		<edge from-layer="402" from-port="0" to-layer="404" to-port="2" />
+		<edge from-layer="403" from-port="0" to-layer="404" to-port="3" />
+		<edge from-layer="404" from-port="5" to-layer="405" to-port="1" />
+		<edge from-layer="405" from-port="2" to-layer="406" to-port="1" />
+		<edge from-layer="406" from-port="2" to-layer="431" to-port="1" />
+		<edge from-layer="406" from-port="2" to-layer="5157" to-port="0" />
+		<edge from-layer="407" from-port="0" to-layer="412" to-port="0" />
+		<edge from-layer="408" from-port="0" to-layer="411" to-port="1" />
+		<edge from-layer="409" from-port="0" to-layer="411" to-port="2" />
 		<edge from-layer="410" from-port="0" to-layer="411" to-port="3" />
 		<edge from-layer="411" from-port="4" to-layer="412" to-port="1" />
-		<edge from-layer="412" from-port="2" to-layer="421" to-port="0" />
-		<edge from-layer="413" from-port="0" to-layer="414" to-port="0" />
-		<edge from-layer="414" from-port="1" to-layer="417" to-port="0" />
-		<edge from-layer="415" from-port="0" to-layer="417" to-port="1" />
-		<edge from-layer="416" from-port="0" to-layer="417" to-port="3" />
-		<edge from-layer="417" from-port="4" to-layer="419" to-port="0" />
-		<edge from-layer="418" from-port="0" to-layer="419" to-port="2" />
-		<edge from-layer="419" from-port="3" to-layer="420" to-port="0" />
-		<edge from-layer="420" from-port="2" to-layer="421" to-port="1" />
-		<edge from-layer="420" from-port="2" to-layer="454" to-port="1" />
+		<edge from-layer="412" from-port="2" to-layer="413" to-port="0" />
+		<edge from-layer="413" from-port="1" to-layer="415" to-port="0" />
+		<edge from-layer="414" from-port="0" to-layer="415" to-port="2" />
+		<edge from-layer="415" from-port="3" to-layer="422" to-port="0" />
+		<edge from-layer="416" from-port="0" to-layer="419" to-port="1" />
+		<edge from-layer="417" from-port="0" to-layer="419" to-port="2" />
+		<edge from-layer="418" from-port="0" to-layer="419" to-port="3" />
+		<edge from-layer="419" from-port="5" to-layer="421" to-port="0" />
+		<edge from-layer="420" from-port="0" to-layer="421" to-port="1" />
 		<edge from-layer="421" from-port="2" to-layer="422" to-port="1" />
-		<edge from-layer="422" from-port="2" to-layer="427" to-port="0" />
-		<edge from-layer="423" from-port="0" to-layer="426" to-port="1" />
-		<edge from-layer="424" from-port="0" to-layer="426" to-port="2" />
-		<edge from-layer="425" from-port="0" to-layer="426" to-port="3" />
-		<edge from-layer="426" from-port="4" to-layer="427" to-port="1" />
-		<edge from-layer="427" from-port="2" to-layer="485" to-port="0" />
-		<edge from-layer="427" from-port="2" to-layer="518" to-port="0" />
-		<edge from-layer="428" from-port="0" to-layer="431" to-port="1" />
-		<edge from-layer="429" from-port="0" to-layer="431" to-port="2" />
-		<edge from-layer="430" from-port="0" to-layer="431" to-port="3" />
-		<edge from-layer="431" from-port="4" to-layer="446" to-port="0" />
-		<edge from-layer="431" from-port="4" to-layer="452" to-port="0" />
-		<edge from-layer="431" from-port="4" to-layer="435" to-port="0" />
-		<edge from-layer="431" from-port="4" to-layer="432" to-port="0" />
-		<edge from-layer="432" from-port="2" to-layer="455" to-port="0" />
-		<edge from-layer="433" from-port="0" to-layer="443" to-port="0" />
-		<edge from-layer="434" from-port="0" to-layer="443" to-port="1" />
-		<edge from-layer="435" from-port="1" to-layer="437" to-port="0" />
-		<edge from-layer="436" from-port="0" to-layer="437" to-port="2" />
-		<edge from-layer="437" from-port="3" to-layer="438" to-port="0" />
-		<edge from-layer="438" from-port="2" to-layer="439" to-port="0" />
-		<edge from-layer="439" from-port="1" to-layer="441" to-port="0" />
-		<edge from-layer="440" from-port="0" to-layer="441" to-port="1" />
-		<edge from-layer="441" from-port="2" to-layer="443" to-port="2" />
-		<edge from-layer="442" from-port="0" to-layer="443" to-port="3" />
-		<edge from-layer="443" from-port="4" to-layer="452" to-port="2" />
-		<edge from-layer="443" from-port="4" to-layer="446" to-port="1" />
-		<edge from-layer="444" from-port="0" to-layer="446" to-port="2" />
-		<edge from-layer="445" from-port="0" to-layer="446" to-port="3" />
-		<edge from-layer="446" from-port="4" to-layer="449" to-port="0" />
-		<edge from-layer="447" from-port="0" to-layer="448" to-port="0" />
-		<edge from-layer="448" from-port="1" to-layer="449" to-port="1" />
-		<edge from-layer="449" from-port="2" to-layer="453" to-port="0" />
-		<edge from-layer="450" from-port="0" to-layer="452" to-port="1" />
-		<edge from-layer="451" from-port="0" to-layer="452" to-port="3" />
-		<edge from-layer="452" from-port="4" to-layer="453" to-port="1" />
-		<edge from-layer="453" from-port="2" to-layer="454" to-port="0" />
-		<edge from-layer="454" from-port="2" to-layer="455" to-port="1" />
-		<edge from-layer="455" from-port="2" to-layer="460" to-port="0" />
-		<edge from-layer="456" from-port="0" to-layer="459" to-port="1" />
-		<edge from-layer="457" from-port="0" to-layer="459" to-port="2" />
-		<edge from-layer="458" from-port="0" to-layer="459" to-port="3" />
-		<edge from-layer="459" from-port="4" to-layer="460" to-port="1" />
-		<edge from-layer="460" from-port="2" to-layer="461" to-port="1" />
-		<edge from-layer="461" from-port="2" to-layer="518" to-port="1" />
-		<edge from-layer="461" from-port="2" to-layer="482" to-port="0" />
-		<edge from-layer="461" from-port="2" to-layer="7720" to-port="0" />
-		<edge from-layer="462" from-port="0" to-layer="463" to-port="0" />
-		<edge from-layer="463" from-port="1" to-layer="468" to-port="0" />
-		<edge from-layer="464" from-port="0" to-layer="467" to-port="1" />
-		<edge from-layer="465" from-port="0" to-layer="467" to-port="2" />
-		<edge from-layer="466" from-port="0" to-layer="467" to-port="3" />
-		<edge from-layer="467" from-port="4" to-layer="468" to-port="1" />
-		<edge from-layer="468" from-port="2" to-layer="469" to-port="0" />
-		<edge from-layer="469" from-port="1" to-layer="471" to-port="0" />
-		<edge from-layer="470" from-port="0" to-layer="471" to-port="2" />
-		<edge from-layer="471" from-port="3" to-layer="478" to-port="0" />
-		<edge from-layer="472" from-port="0" to-layer="475" to-port="1" />
-		<edge from-layer="473" from-port="0" to-layer="475" to-port="2" />
-		<edge from-layer="474" from-port="0" to-layer="475" to-port="3" />
-		<edge from-layer="475" from-port="4" to-layer="477" to-port="0" />
-		<edge from-layer="476" from-port="0" to-layer="477" to-port="1" />
-		<edge from-layer="477" from-port="2" to-layer="478" to-port="1" />
-		<edge from-layer="478" from-port="2" to-layer="7719" to-port="0" />
-		<edge from-layer="478" from-port="2" to-layer="518" to-port="2" />
-		<edge from-layer="479" from-port="0" to-layer="497" to-port="0" />
-		<edge from-layer="480" from-port="0" to-layer="492" to-port="0" />
-		<edge from-layer="481" from-port="0" to-layer="495" to-port="1" />
-		<edge from-layer="481" from-port="0" to-layer="492" to-port="1" />
-		<edge from-layer="482" from-port="1" to-layer="484" to-port="0" />
-		<edge from-layer="483" from-port="0" to-layer="484" to-port="2" />
-		<edge from-layer="484" from-port="3" to-layer="488" to-port="0" />
-		<edge from-layer="484" from-port="3" to-layer="494" to-port="0" />
-		<edge from-layer="485" from-port="1" to-layer="487" to-port="0" />
-		<edge from-layer="485" from-port="1" to-layer="513" to-port="0" />
-		<edge from-layer="486" from-port="0" to-layer="487" to-port="2" />
-		<edge from-layer="487" from-port="3" to-layer="488" to-port="1" />
+		<edge from-layer="422" from-port="2" to-layer="431" to-port="2" />
+		<edge from-layer="422" from-port="2" to-layer="5156" to-port="0" />
+		<edge from-layer="423" from-port="0" to-layer="430" to-port="1" />
+		<edge from-layer="424" from-port="1" to-layer="427" to-port="0" />
+		<edge from-layer="425" from-port="0" to-layer="427" to-port="1" />
+		<edge from-layer="426" from-port="0" to-layer="427" to-port="2" />
+		<edge from-layer="427" from-port="3" to-layer="428" to-port="0" />
+		<edge from-layer="428" from-port="2" to-layer="430" to-port="2" />
+		<edge from-layer="429" from-port="0" to-layer="430" to-port="3" />
+		<edge from-layer="430" from-port="5" to-layer="431" to-port="3" />
+		<edge from-layer="431" from-port="4" to-layer="433" to-port="0" />
+		<edge from-layer="432" from-port="0" to-layer="433" to-port="1" />
+		<edge from-layer="433" from-port="2" to-layer="435" to-port="0" />
+		<edge from-layer="434" from-port="0" to-layer="435" to-port="1" />
+		<edge from-layer="435" from-port="2" to-layer="444" to-port="0" />
+		<edge from-layer="436" from-port="0" to-layer="437" to-port="0" />
+		<edge from-layer="437" from-port="1" to-layer="440" to-port="0" />
+		<edge from-layer="438" from-port="0" to-layer="439" to-port="0" />
+		<edge from-layer="439" from-port="1" to-layer="440" to-port="1" />
+		<edge from-layer="440" from-port="2" to-layer="442" to-port="0" />
+		<edge from-layer="441" from-port="0" to-layer="442" to-port="1" />
+		<edge from-layer="442" from-port="2" to-layer="443" to-port="0" />
+		<edge from-layer="443" from-port="1" to-layer="444" to-port="1" />
+		<edge from-layer="444" from-port="2" to-layer="446" to-port="0" />
+		<edge from-layer="445" from-port="0" to-layer="446" to-port="1" />
+		<edge from-layer="446" from-port="2" to-layer="447" to-port="1" />
+		<edge from-layer="447" from-port="2" to-layer="448" to-port="0" />
+		<edge from-layer="448" from-port="2" to-layer="606" to-port="1" />
+		<edge from-layer="448" from-port="2" to-layer="450" to-port="0" />
+		<edge from-layer="449" from-port="0" to-layer="450" to-port="1" />
+		<edge from-layer="450" from-port="2" to-layer="481" to-port="0" />
+		<edge from-layer="450" from-port="2" to-layer="452" to-port="0" />
+		<edge from-layer="451" from-port="0" to-layer="452" to-port="1" />
+		<edge from-layer="452" from-port="2" to-layer="454" to-port="0" />
+		<edge from-layer="453" from-port="0" to-layer="454" to-port="1" />
+		<edge from-layer="454" from-port="2" to-layer="463" to-port="0" />
+		<edge from-layer="455" from-port="0" to-layer="456" to-port="0" />
+		<edge from-layer="456" from-port="1" to-layer="459" to-port="0" />
+		<edge from-layer="457" from-port="0" to-layer="458" to-port="0" />
+		<edge from-layer="458" from-port="1" to-layer="459" to-port="1" />
+		<edge from-layer="459" from-port="2" to-layer="461" to-port="0" />
+		<edge from-layer="460" from-port="0" to-layer="461" to-port="1" />
+		<edge from-layer="461" from-port="2" to-layer="462" to-port="0" />
+		<edge from-layer="462" from-port="1" to-layer="463" to-port="1" />
+		<edge from-layer="463" from-port="2" to-layer="466" to-port="0" />
+		<edge from-layer="464" from-port="0" to-layer="465" to-port="0" />
+		<edge from-layer="465" from-port="1" to-layer="466" to-port="1" />
+		<edge from-layer="466" from-port="2" to-layer="467" to-port="0" />
+		<edge from-layer="467" from-port="1" to-layer="476" to-port="0" />
+		<edge from-layer="468" from-port="0" to-layer="469" to-port="0" />
+		<edge from-layer="469" from-port="1" to-layer="472" to-port="0" />
+		<edge from-layer="470" from-port="0" to-layer="471" to-port="0" />
+		<edge from-layer="471" from-port="1" to-layer="472" to-port="1" />
+		<edge from-layer="472" from-port="2" to-layer="474" to-port="0" />
+		<edge from-layer="473" from-port="0" to-layer="474" to-port="1" />
+		<edge from-layer="474" from-port="2" to-layer="475" to-port="0" />
+		<edge from-layer="475" from-port="1" to-layer="476" to-port="1" />
+		<edge from-layer="476" from-port="2" to-layer="479" to-port="0" />
+		<edge from-layer="477" from-port="0" to-layer="478" to-port="0" />
+		<edge from-layer="478" from-port="1" to-layer="479" to-port="1" />
+		<edge from-layer="479" from-port="2" to-layer="605" to-port="0" />
+		<edge from-layer="480" from-port="0" to-layer="481" to-port="1" />
+		<edge from-layer="481" from-port="2" to-layer="483" to-port="0" />
+		<edge from-layer="482" from-port="0" to-layer="483" to-port="1" />
+		<edge from-layer="483" from-port="2" to-layer="492" to-port="0" />
+		<edge from-layer="484" from-port="0" to-layer="485" to-port="0" />
+		<edge from-layer="485" from-port="1" to-layer="488" to-port="0" />
+		<edge from-layer="486" from-port="0" to-layer="487" to-port="0" />
+		<edge from-layer="487" from-port="1" to-layer="488" to-port="1" />
 		<edge from-layer="488" from-port="2" to-layer="490" to-port="0" />
-		<edge from-layer="489" from-port="0" to-layer="494" to-port="1" />
 		<edge from-layer="489" from-port="0" to-layer="490" to-port="1" />
-		<edge from-layer="490" from-port="2" to-layer="492" to-port="2" />
-		<edge from-layer="491" from-port="0" to-layer="492" to-port="3" />
-		<edge from-layer="491" from-port="0" to-layer="495" to-port="3" />
-		<edge from-layer="492" from-port="4" to-layer="497" to-port="1" />
-		<edge from-layer="493" from-port="0" to-layer="495" to-port="0" />
-		<edge from-layer="494" from-port="2" to-layer="502" to-port="2" />
-		<edge from-layer="494" from-port="2" to-layer="495" to-port="2" />
-		<edge from-layer="495" from-port="4" to-layer="497" to-port="2" />
-		<edge from-layer="496" from-port="0" to-layer="497" to-port="3" />
-		<edge from-layer="497" from-port="4" to-layer="504" to-port="0" />
-		<edge from-layer="498" from-port="0" to-layer="504" to-port="1" />
-		<edge from-layer="499" from-port="0" to-layer="502" to-port="0" />
-		<edge from-layer="500" from-port="0" to-layer="502" to-port="1" />
-		<edge from-layer="501" from-port="0" to-layer="502" to-port="3" />
-		<edge from-layer="502" from-port="4" to-layer="504" to-port="2" />
-		<edge from-layer="503" from-port="0" to-layer="504" to-port="3" />
-		<edge from-layer="504" from-port="4" to-layer="508" to-port="0" />
-		<edge from-layer="505" from-port="0" to-layer="506" to-port="0" />
-		<edge from-layer="506" from-port="1" to-layer="508" to-port="1" />
-		<edge from-layer="507" from-port="0" to-layer="508" to-port="2" />
-		<edge from-layer="508" from-port="3" to-layer="509" to-port="0" />
-		<edge from-layer="509" from-port="2" to-layer="517" to-port="0" />
-		<edge from-layer="509" from-port="2" to-layer="510" to-port="0" />
-		<edge from-layer="510" from-port="1" to-layer="516" to-port="0" />
-		<edge from-layer="511" from-port="0" to-layer="513" to-port="1" />
-		<edge from-layer="512" from-port="0" to-layer="513" to-port="2" />
-		<edge from-layer="513" from-port="3" to-layer="514" to-port="0" />
-		<edge from-layer="514" from-port="4" to-layer="515" to-port="0" />
-		<edge from-layer="515" from-port="1" to-layer="516" to-port="1" />
-		<edge from-layer="516" from-port="2" to-layer="517" to-port="1" />
-		<edge from-layer="517" from-port="2" to-layer="518" to-port="3" />
-		<edge from-layer="518" from-port="4" to-layer="520" to-port="0" />
-		<edge from-layer="519" from-port="0" to-layer="520" to-port="1" />
-		<edge from-layer="520" from-port="2" to-layer="522" to-port="0" />
-		<edge from-layer="521" from-port="0" to-layer="522" to-port="1" />
-		<edge from-layer="522" from-port="2" to-layer="531" to-port="0" />
-		<edge from-layer="523" from-port="0" to-layer="524" to-port="0" />
-		<edge from-layer="524" from-port="1" to-layer="527" to-port="0" />
-		<edge from-layer="525" from-port="0" to-layer="526" to-port="0" />
-		<edge from-layer="526" from-port="1" to-layer="527" to-port="1" />
-		<edge from-layer="527" from-port="2" to-layer="529" to-port="0" />
-		<edge from-layer="528" from-port="0" to-layer="529" to-port="1" />
-		<edge from-layer="529" from-port="2" to-layer="530" to-port="0" />
-		<edge from-layer="530" from-port="1" to-layer="531" to-port="1" />
-		<edge from-layer="531" from-port="2" to-layer="534" to-port="0" />
-		<edge from-layer="532" from-port="0" to-layer="533" to-port="0" />
-		<edge from-layer="533" from-port="1" to-layer="534" to-port="1" />
-		<edge from-layer="534" from-port="2" to-layer="535" to-port="1" />
-		<edge from-layer="535" from-port="2" to-layer="536" to-port="0" />
-		<edge from-layer="536" from-port="2" to-layer="778" to-port="1" />
-		<edge from-layer="536" from-port="2" to-layer="538" to-port="0" />
-		<edge from-layer="537" from-port="0" to-layer="538" to-port="1" />
-		<edge from-layer="538" from-port="2" to-layer="572" to-port="0" />
-		<edge from-layer="538" from-port="2" to-layer="541" to-port="0" />
-		<edge from-layer="539" from-port="0" to-layer="540" to-port="0" />
-		<edge from-layer="540" from-port="1" to-layer="541" to-port="1" />
-		<edge from-layer="541" from-port="2" to-layer="544" to-port="0" />
-		<edge from-layer="542" from-port="0" to-layer="543" to-port="0" />
-		<edge from-layer="543" from-port="1" to-layer="544" to-port="1" />
-		<edge from-layer="544" from-port="2" to-layer="553" to-port="0" />
-		<edge from-layer="545" from-port="0" to-layer="546" to-port="0" />
-		<edge from-layer="546" from-port="1" to-layer="549" to-port="0" />
-		<edge from-layer="547" from-port="0" to-layer="548" to-port="0" />
-		<edge from-layer="548" from-port="1" to-layer="549" to-port="1" />
-		<edge from-layer="549" from-port="2" to-layer="551" to-port="0" />
+		<edge from-layer="490" from-port="2" to-layer="491" to-port="0" />
+		<edge from-layer="491" from-port="1" to-layer="492" to-port="1" />
+		<edge from-layer="492" from-port="2" to-layer="494" to-port="0" />
+		<edge from-layer="493" from-port="0" to-layer="494" to-port="1" />
+		<edge from-layer="494" from-port="2" to-layer="496" to-port="0" />
+		<edge from-layer="495" from-port="0" to-layer="496" to-port="1" />
+		<edge from-layer="496" from-port="2" to-layer="500" to-port="0" />
+		<edge from-layer="496" from-port="2" to-layer="577" to-port="0" />
+		<edge from-layer="496" from-port="2" to-layer="538" to-port="0" />
+		<edge from-layer="497" from-port="0" to-layer="500" to-port="1" />
+		<edge from-layer="498" from-port="0" to-layer="500" to-port="2" />
+		<edge from-layer="499" from-port="0" to-layer="500" to-port="3" />
+		<edge from-layer="500" from-port="5" to-layer="502" to-port="0" />
+		<edge from-layer="501" from-port="0" to-layer="502" to-port="1" />
+		<edge from-layer="502" from-port="2" to-layer="506" to-port="0" />
+		<edge from-layer="502" from-port="2" to-layer="524" to-port="0" />
+		<edge from-layer="503" from-port="0" to-layer="506" to-port="1" />
+		<edge from-layer="504" from-port="0" to-layer="506" to-port="2" />
+		<edge from-layer="505" from-port="0" to-layer="506" to-port="3" />
+		<edge from-layer="506" from-port="5" to-layer="511" to-port="0" />
+		<edge from-layer="506" from-port="5" to-layer="517" to-port="0" />
+		<edge from-layer="506" from-port="5" to-layer="507" to-port="0" />
+		<edge from-layer="507" from-port="2" to-layer="520" to-port="0" />
+		<edge from-layer="508" from-port="0" to-layer="511" to-port="1" />
+		<edge from-layer="509" from-port="0" to-layer="511" to-port="2" />
+		<edge from-layer="510" from-port="0" to-layer="511" to-port="3" />
+		<edge from-layer="511" from-port="5" to-layer="513" to-port="0" />
+		<edge from-layer="512" from-port="0" to-layer="513" to-port="1" />
+		<edge from-layer="513" from-port="2" to-layer="518" to-port="0" />
+		<edge from-layer="514" from-port="0" to-layer="517" to-port="1" />
+		<edge from-layer="515" from-port="0" to-layer="517" to-port="2" />
+		<edge from-layer="516" from-port="0" to-layer="517" to-port="3" />
+		<edge from-layer="517" from-port="5" to-layer="518" to-port="1" />
+		<edge from-layer="518" from-port="2" to-layer="519" to-port="0" />
+		<edge from-layer="519" from-port="2" to-layer="520" to-port="1" />
+		<edge from-layer="520" from-port="2" to-layer="525" to-port="0" />
+		<edge from-layer="521" from-port="0" to-layer="524" to-port="1" />
+		<edge from-layer="522" from-port="0" to-layer="524" to-port="2" />
+		<edge from-layer="523" from-port="0" to-layer="524" to-port="3" />
+		<edge from-layer="524" from-port="5" to-layer="525" to-port="1" />
+		<edge from-layer="525" from-port="2" to-layer="589" to-port="0" />
+		<edge from-layer="526" from-port="0" to-layer="531" to-port="0" />
+		<edge from-layer="527" from-port="0" to-layer="530" to-port="1" />
+		<edge from-layer="528" from-port="0" to-layer="530" to-port="2" />
+		<edge from-layer="529" from-port="0" to-layer="530" to-port="3" />
+		<edge from-layer="530" from-port="4" to-layer="531" to-port="1" />
+		<edge from-layer="531" from-port="2" to-layer="532" to-port="0" />
+		<edge from-layer="532" from-port="1" to-layer="534" to-port="0" />
+		<edge from-layer="533" from-port="0" to-layer="534" to-port="2" />
+		<edge from-layer="534" from-port="3" to-layer="564" to-port="0" />
+		<edge from-layer="534" from-port="3" to-layer="582" to-port="0" />
+		<edge from-layer="535" from-port="0" to-layer="538" to-port="1" />
+		<edge from-layer="536" from-port="0" to-layer="538" to-port="2" />
+		<edge from-layer="537" from-port="0" to-layer="538" to-port="3" />
+		<edge from-layer="538" from-port="5" to-layer="540" to-port="0" />
+		<edge from-layer="539" from-port="0" to-layer="540" to-port="1" />
+		<edge from-layer="540" from-port="2" to-layer="562" to-port="0" />
+		<edge from-layer="540" from-port="2" to-layer="544" to-port="0" />
+		<edge from-layer="541" from-port="0" to-layer="544" to-port="1" />
+		<edge from-layer="542" from-port="0" to-layer="544" to-port="2" />
+		<edge from-layer="543" from-port="0" to-layer="544" to-port="3" />
+		<edge from-layer="544" from-port="5" to-layer="545" to-port="0" />
+		<edge from-layer="544" from-port="5" to-layer="549" to-port="0" />
+		<edge from-layer="544" from-port="5" to-layer="555" to-port="0" />
+		<edge from-layer="545" from-port="2" to-layer="558" to-port="0" />
+		<edge from-layer="546" from-port="0" to-layer="549" to-port="1" />
+		<edge from-layer="547" from-port="0" to-layer="549" to-port="2" />
+		<edge from-layer="548" from-port="0" to-layer="549" to-port="3" />
+		<edge from-layer="549" from-port="5" to-layer="551" to-port="0" />
 		<edge from-layer="550" from-port="0" to-layer="551" to-port="1" />
-		<edge from-layer="551" from-port="2" to-layer="552" to-port="0" />
-		<edge from-layer="552" from-port="1" to-layer="553" to-port="1" />
-		<edge from-layer="553" from-port="2" to-layer="556" to-port="0" />
-		<edge from-layer="554" from-port="0" to-layer="555" to-port="0" />
-		<edge from-layer="555" from-port="1" to-layer="556" to-port="1" />
+		<edge from-layer="551" from-port="2" to-layer="556" to-port="0" />
+		<edge from-layer="552" from-port="0" to-layer="555" to-port="1" />
+		<edge from-layer="553" from-port="0" to-layer="555" to-port="2" />
+		<edge from-layer="554" from-port="0" to-layer="555" to-port="3" />
+		<edge from-layer="555" from-port="5" to-layer="556" to-port="1" />
 		<edge from-layer="556" from-port="2" to-layer="557" to-port="0" />
-		<edge from-layer="557" from-port="1" to-layer="566" to-port="0" />
-		<edge from-layer="558" from-port="0" to-layer="559" to-port="0" />
-		<edge from-layer="559" from-port="1" to-layer="562" to-port="0" />
-		<edge from-layer="560" from-port="0" to-layer="561" to-port="0" />
-		<edge from-layer="561" from-port="1" to-layer="562" to-port="1" />
-		<edge from-layer="562" from-port="2" to-layer="564" to-port="0" />
-		<edge from-layer="563" from-port="0" to-layer="564" to-port="1" />
-		<edge from-layer="564" from-port="2" to-layer="565" to-port="0" />
-		<edge from-layer="565" from-port="1" to-layer="566" to-port="1" />
-		<edge from-layer="566" from-port="2" to-layer="569" to-port="0" />
-		<edge from-layer="567" from-port="0" to-layer="568" to-port="0" />
-		<edge from-layer="568" from-port="1" to-layer="569" to-port="1" />
-		<edge from-layer="569" from-port="2" to-layer="777" to-port="0" />
-		<edge from-layer="570" from-port="0" to-layer="571" to-port="0" />
-		<edge from-layer="571" from-port="1" to-layer="572" to-port="1" />
-		<edge from-layer="572" from-port="2" to-layer="575" to-port="0" />
-		<edge from-layer="573" from-port="0" to-layer="574" to-port="0" />
-		<edge from-layer="574" from-port="1" to-layer="575" to-port="1" />
-		<edge from-layer="575" from-port="2" to-layer="584" to-port="0" />
-		<edge from-layer="576" from-port="0" to-layer="577" to-port="0" />
-		<edge from-layer="577" from-port="1" to-layer="580" to-port="0" />
-		<edge from-layer="578" from-port="0" to-layer="579" to-port="0" />
-		<edge from-layer="579" from-port="1" to-layer="580" to-port="1" />
-		<edge from-layer="580" from-port="2" to-layer="582" to-port="0" />
-		<edge from-layer="581" from-port="0" to-layer="582" to-port="1" />
-		<edge from-layer="582" from-port="2" to-layer="583" to-port="0" />
-		<edge from-layer="583" from-port="1" to-layer="584" to-port="1" />
-		<edge from-layer="584" from-port="2" to-layer="587" to-port="0" />
-		<edge from-layer="585" from-port="0" to-layer="586" to-port="0" />
-		<edge from-layer="586" from-port="1" to-layer="587" to-port="1" />
-		<edge from-layer="587" from-port="2" to-layer="589" to-port="0" />
-		<edge from-layer="588" from-port="0" to-layer="589" to-port="1" />
-		<edge from-layer="589" from-port="2" to-layer="593" to-port="0" />
-		<edge from-layer="589" from-port="2" to-layer="717" to-port="0" />
-		<edge from-layer="589" from-port="2" to-layer="606" to-port="0" />
-		<edge from-layer="590" from-port="0" to-layer="593" to-port="1" />
-		<edge from-layer="591" from-port="0" to-layer="593" to-port="2" />
-		<edge from-layer="592" from-port="0" to-layer="593" to-port="3" />
-		<edge from-layer="593" from-port="4" to-layer="595" to-port="0" />
-		<edge from-layer="594" from-port="0" to-layer="595" to-port="1" />
-		<edge from-layer="595" from-port="2" to-layer="668" to-port="0" />
-		<edge from-layer="595" from-port="2" to-layer="599" to-port="0" />
-		<edge from-layer="596" from-port="0" to-layer="599" to-port="1" />
-		<edge from-layer="597" from-port="0" to-layer="599" to-port="2" />
-		<edge from-layer="598" from-port="0" to-layer="599" to-port="3" />
-		<edge from-layer="599" from-port="4" to-layer="647" to-port="0" />
-		<edge from-layer="599" from-port="4" to-layer="653" to-port="0" />
-		<edge from-layer="599" from-port="4" to-layer="636" to-port="0" />
-		<edge from-layer="599" from-port="4" to-layer="633" to-port="0" />
-		<edge from-layer="600" from-port="0" to-layer="601" to-port="0" />
-		<edge from-layer="601" from-port="1" to-layer="629" to-port="0" />
-		<edge from-layer="602" from-port="0" to-layer="629" to-port="1" />
-		<edge from-layer="603" from-port="0" to-layer="606" to-port="1" />
-		<edge from-layer="604" from-port="0" to-layer="606" to-port="2" />
-		<edge from-layer="605" from-port="0" to-layer="606" to-port="3" />
-		<edge from-layer="606" from-port="4" to-layer="608" to-port="0" />
+		<edge from-layer="557" from-port="2" to-layer="558" to-port="1" />
+		<edge from-layer="558" from-port="2" to-layer="563" to-port="0" />
+		<edge from-layer="559" from-port="0" to-layer="562" to-port="1" />
+		<edge from-layer="560" from-port="0" to-layer="562" to-port="2" />
+		<edge from-layer="561" from-port="0" to-layer="562" to-port="3" />
+		<edge from-layer="562" from-port="5" to-layer="563" to-port="1" />
+		<edge from-layer="563" from-port="2" to-layer="564" to-port="1" />
+		<edge from-layer="564" from-port="2" to-layer="589" to-port="1" />
+		<edge from-layer="564" from-port="2" to-layer="5135" to-port="0" />
+		<edge from-layer="565" from-port="0" to-layer="570" to-port="0" />
+		<edge from-layer="566" from-port="0" to-layer="569" to-port="1" />
+		<edge from-layer="567" from-port="0" to-layer="569" to-port="2" />
+		<edge from-layer="568" from-port="0" to-layer="569" to-port="3" />
+		<edge from-layer="569" from-port="4" to-layer="570" to-port="1" />
+		<edge from-layer="570" from-port="2" to-layer="571" to-port="0" />
+		<edge from-layer="571" from-port="1" to-layer="573" to-port="0" />
+		<edge from-layer="572" from-port="0" to-layer="573" to-port="2" />
+		<edge from-layer="573" from-port="3" to-layer="580" to-port="0" />
+		<edge from-layer="574" from-port="0" to-layer="577" to-port="1" />
+		<edge from-layer="575" from-port="0" to-layer="577" to-port="2" />
+		<edge from-layer="576" from-port="0" to-layer="577" to-port="3" />
+		<edge from-layer="577" from-port="5" to-layer="579" to-port="0" />
+		<edge from-layer="578" from-port="0" to-layer="579" to-port="1" />
+		<edge from-layer="579" from-port="2" to-layer="580" to-port="1" />
+		<edge from-layer="580" from-port="2" to-layer="589" to-port="2" />
+		<edge from-layer="580" from-port="2" to-layer="5134" to-port="0" />
+		<edge from-layer="581" from-port="0" to-layer="588" to-port="1" />
+		<edge from-layer="582" from-port="1" to-layer="585" to-port="0" />
+		<edge from-layer="583" from-port="0" to-layer="585" to-port="1" />
+		<edge from-layer="584" from-port="0" to-layer="585" to-port="2" />
+		<edge from-layer="585" from-port="3" to-layer="586" to-port="0" />
+		<edge from-layer="586" from-port="2" to-layer="588" to-port="2" />
+		<edge from-layer="587" from-port="0" to-layer="588" to-port="3" />
+		<edge from-layer="588" from-port="5" to-layer="589" to-port="3" />
+		<edge from-layer="589" from-port="4" to-layer="591" to-port="0" />
+		<edge from-layer="590" from-port="0" to-layer="591" to-port="1" />
+		<edge from-layer="591" from-port="2" to-layer="593" to-port="0" />
+		<edge from-layer="592" from-port="0" to-layer="593" to-port="1" />
+		<edge from-layer="593" from-port="2" to-layer="602" to-port="0" />
+		<edge from-layer="594" from-port="0" to-layer="595" to-port="0" />
+		<edge from-layer="595" from-port="1" to-layer="598" to-port="0" />
+		<edge from-layer="596" from-port="0" to-layer="597" to-port="0" />
+		<edge from-layer="597" from-port="1" to-layer="598" to-port="1" />
+		<edge from-layer="598" from-port="2" to-layer="600" to-port="0" />
+		<edge from-layer="599" from-port="0" to-layer="600" to-port="1" />
+		<edge from-layer="600" from-port="2" to-layer="601" to-port="0" />
+		<edge from-layer="601" from-port="1" to-layer="602" to-port="1" />
+		<edge from-layer="602" from-port="2" to-layer="604" to-port="0" />
+		<edge from-layer="603" from-port="0" to-layer="604" to-port="1" />
+		<edge from-layer="604" from-port="2" to-layer="605" to-port="1" />
+		<edge from-layer="605" from-port="2" to-layer="606" to-port="0" />
+		<edge from-layer="606" from-port="2" to-layer="608" to-port="0" />
+		<edge from-layer="606" from-port="2" to-layer="764" to-port="1" />
 		<edge from-layer="607" from-port="0" to-layer="608" to-port="1" />
-		<edge from-layer="608" from-port="2" to-layer="609" to-port="0" />
-		<edge from-layer="608" from-port="2" to-layer="673" to-port="0" />
-		<edge from-layer="608" from-port="2" to-layer="701" to-port="0" />
-		<edge from-layer="609" from-port="1" to-layer="611" to-port="0" />
-		<edge from-layer="610" from-port="0" to-layer="611" to-port="2" />
-		<edge from-layer="611" from-port="3" to-layer="625" to-port="0" />
-		<edge from-layer="612" from-port="0" to-layer="613" to-port="0" />
-		<edge from-layer="613" from-port="1" to-layer="618" to-port="0" />
-		<edge from-layer="614" from-port="0" to-layer="617" to-port="1" />
-		<edge from-layer="615" from-port="0" to-layer="617" to-port="2" />
-		<edge from-layer="616" from-port="0" to-layer="617" to-port="3" />
-		<edge from-layer="617" from-port="4" to-layer="618" to-port="1" />
-		<edge from-layer="618" from-port="2" to-layer="619" to-port="0" />
-		<edge from-layer="619" from-port="1" to-layer="621" to-port="0" />
-		<edge from-layer="620" from-port="0" to-layer="621" to-port="2" />
-		<edge from-layer="621" from-port="3" to-layer="622" to-port="0" />
-		<edge from-layer="621" from-port="3" to-layer="703" to-port="0" />
-		<edge from-layer="622" from-port="1" to-layer="624" to-port="0" />
-		<edge from-layer="623" from-port="0" to-layer="624" to-port="2" />
-		<edge from-layer="624" from-port="3" to-layer="625" to-port="1" />
-		<edge from-layer="625" from-port="2" to-layer="627" to-port="0" />
-		<edge from-layer="626" from-port="0" to-layer="627" to-port="1" />
-		<edge from-layer="627" from-port="2" to-layer="629" to-port="2" />
-		<edge from-layer="627" from-port="2" to-layer="659" to-port="2" />
-		<edge from-layer="628" from-port="0" to-layer="629" to-port="3" />
-		<edge from-layer="629" from-port="4" to-layer="631" to-port="0" />
-		<edge from-layer="630" from-port="0" to-layer="631" to-port="2" />
-		<edge from-layer="631" from-port="3" to-layer="632" to-port="0" />
-		<edge from-layer="632" from-port="2" to-layer="674" to-port="1" />
-		<edge from-layer="632" from-port="2" to-layer="633" to-port="1" />
-		<edge from-layer="633" from-port="2" to-layer="664" to-port="0" />
-		<edge from-layer="634" from-port="0" to-layer="644" to-port="0" />
-		<edge from-layer="635" from-port="0" to-layer="644" to-port="1" />
-		<edge from-layer="636" from-port="1" to-layer="638" to-port="0" />
-		<edge from-layer="637" from-port="0" to-layer="638" to-port="2" />
-		<edge from-layer="638" from-port="3" to-layer="639" to-port="0" />
-		<edge from-layer="639" from-port="2" to-layer="640" to-port="0" />
-		<edge from-layer="640" from-port="1" to-layer="642" to-port="0" />
-		<edge from-layer="641" from-port="0" to-layer="642" to-port="1" />
-		<edge from-layer="642" from-port="2" to-layer="644" to-port="2" />
-		<edge from-layer="643" from-port="0" to-layer="644" to-port="3" />
-		<edge from-layer="644" from-port="4" to-layer="653" to-port="2" />
-		<edge from-layer="644" from-port="4" to-layer="647" to-port="1" />
-		<edge from-layer="645" from-port="0" to-layer="647" to-port="2" />
-		<edge from-layer="646" from-port="0" to-layer="647" to-port="3" />
-		<edge from-layer="647" from-port="4" to-layer="650" to-port="0" />
-		<edge from-layer="648" from-port="0" to-layer="649" to-port="0" />
+		<edge from-layer="608" from-port="2" to-layer="639" to-port="0" />
+		<edge from-layer="608" from-port="2" to-layer="610" to-port="0" />
+		<edge from-layer="609" from-port="0" to-layer="610" to-port="1" />
+		<edge from-layer="610" from-port="2" to-layer="612" to-port="0" />
+		<edge from-layer="611" from-port="0" to-layer="612" to-port="1" />
+		<edge from-layer="612" from-port="2" to-layer="621" to-port="0" />
+		<edge from-layer="613" from-port="0" to-layer="614" to-port="0" />
+		<edge from-layer="614" from-port="1" to-layer="617" to-port="0" />
+		<edge from-layer="615" from-port="0" to-layer="616" to-port="0" />
+		<edge from-layer="616" from-port="1" to-layer="617" to-port="1" />
+		<edge from-layer="617" from-port="2" to-layer="619" to-port="0" />
+		<edge from-layer="618" from-port="0" to-layer="619" to-port="1" />
+		<edge from-layer="619" from-port="2" to-layer="620" to-port="0" />
+		<edge from-layer="620" from-port="1" to-layer="621" to-port="1" />
+		<edge from-layer="621" from-port="2" to-layer="624" to-port="0" />
+		<edge from-layer="622" from-port="0" to-layer="623" to-port="0" />
+		<edge from-layer="623" from-port="1" to-layer="624" to-port="1" />
+		<edge from-layer="624" from-port="2" to-layer="625" to-port="0" />
+		<edge from-layer="625" from-port="1" to-layer="634" to-port="0" />
+		<edge from-layer="626" from-port="0" to-layer="627" to-port="0" />
+		<edge from-layer="627" from-port="1" to-layer="630" to-port="0" />
+		<edge from-layer="628" from-port="0" to-layer="629" to-port="0" />
+		<edge from-layer="629" from-port="1" to-layer="630" to-port="1" />
+		<edge from-layer="630" from-port="2" to-layer="632" to-port="0" />
+		<edge from-layer="631" from-port="0" to-layer="632" to-port="1" />
+		<edge from-layer="632" from-port="2" to-layer="633" to-port="0" />
+		<edge from-layer="633" from-port="1" to-layer="634" to-port="1" />
+		<edge from-layer="634" from-port="2" to-layer="637" to-port="0" />
+		<edge from-layer="635" from-port="0" to-layer="636" to-port="0" />
+		<edge from-layer="636" from-port="1" to-layer="637" to-port="1" />
+		<edge from-layer="637" from-port="2" to-layer="763" to-port="0" />
+		<edge from-layer="638" from-port="0" to-layer="639" to-port="1" />
+		<edge from-layer="639" from-port="2" to-layer="641" to-port="0" />
+		<edge from-layer="640" from-port="0" to-layer="641" to-port="1" />
+		<edge from-layer="641" from-port="2" to-layer="650" to-port="0" />
+		<edge from-layer="642" from-port="0" to-layer="643" to-port="0" />
+		<edge from-layer="643" from-port="1" to-layer="646" to-port="0" />
+		<edge from-layer="644" from-port="0" to-layer="645" to-port="0" />
+		<edge from-layer="645" from-port="1" to-layer="646" to-port="1" />
+		<edge from-layer="646" from-port="2" to-layer="648" to-port="0" />
+		<edge from-layer="647" from-port="0" to-layer="648" to-port="1" />
+		<edge from-layer="648" from-port="2" to-layer="649" to-port="0" />
 		<edge from-layer="649" from-port="1" to-layer="650" to-port="1" />
-		<edge from-layer="650" from-port="2" to-layer="654" to-port="0" />
-		<edge from-layer="651" from-port="0" to-layer="653" to-port="1" />
-		<edge from-layer="652" from-port="0" to-layer="653" to-port="3" />
-		<edge from-layer="653" from-port="4" to-layer="654" to-port="1" />
-		<edge from-layer="654" from-port="2" to-layer="663" to-port="0" />
-		<edge from-layer="655" from-port="0" to-layer="656" to-port="0" />
-		<edge from-layer="656" from-port="1" to-layer="659" to-port="0" />
-		<edge from-layer="657" from-port="0" to-layer="659" to-port="1" />
-		<edge from-layer="658" from-port="0" to-layer="659" to-port="3" />
-		<edge from-layer="659" from-port="4" to-layer="661" to-port="0" />
-		<edge from-layer="660" from-port="0" to-layer="661" to-port="2" />
-		<edge from-layer="661" from-port="3" to-layer="662" to-port="0" />
-		<edge from-layer="662" from-port="2" to-layer="663" to-port="1" />
-		<edge from-layer="662" from-port="2" to-layer="696" to-port="1" />
-		<edge from-layer="663" from-port="2" to-layer="664" to-port="1" />
-		<edge from-layer="664" from-port="2" to-layer="669" to-port="0" />
-		<edge from-layer="665" from-port="0" to-layer="668" to-port="1" />
-		<edge from-layer="666" from-port="0" to-layer="668" to-port="2" />
-		<edge from-layer="667" from-port="0" to-layer="668" to-port="3" />
-		<edge from-layer="668" from-port="4" to-layer="669" to-port="1" />
-		<edge from-layer="669" from-port="2" to-layer="760" to-port="0" />
-		<edge from-layer="669" from-port="2" to-layer="727" to-port="0" />
-		<edge from-layer="670" from-port="0" to-layer="673" to-port="1" />
-		<edge from-layer="671" from-port="0" to-layer="673" to-port="2" />
-		<edge from-layer="672" from-port="0" to-layer="673" to-port="3" />
-		<edge from-layer="673" from-port="4" to-layer="674" to-port="0" />
-		<edge from-layer="673" from-port="4" to-layer="688" to-port="0" />
-		<edge from-layer="673" from-port="4" to-layer="677" to-port="0" />
-		<edge from-layer="673" from-port="4" to-layer="694" to-port="0" />
-		<edge from-layer="674" from-port="2" to-layer="697" to-port="0" />
-		<edge from-layer="675" from-port="0" to-layer="685" to-port="0" />
-		<edge from-layer="676" from-port="0" to-layer="685" to-port="1" />
-		<edge from-layer="677" from-port="1" to-layer="679" to-port="0" />
-		<edge from-layer="678" from-port="0" to-layer="679" to-port="2" />
-		<edge from-layer="679" from-port="3" to-layer="680" to-port="0" />
-		<edge from-layer="680" from-port="2" to-layer="681" to-port="0" />
-		<edge from-layer="681" from-port="1" to-layer="683" to-port="0" />
-		<edge from-layer="682" from-port="0" to-layer="683" to-port="1" />
-		<edge from-layer="683" from-port="2" to-layer="685" to-port="2" />
-		<edge from-layer="684" from-port="0" to-layer="685" to-port="3" />
-		<edge from-layer="685" from-port="4" to-layer="694" to-port="2" />
-		<edge from-layer="685" from-port="4" to-layer="688" to-port="1" />
+		<edge from-layer="650" from-port="2" to-layer="652" to-port="0" />
+		<edge from-layer="651" from-port="0" to-layer="652" to-port="1" />
+		<edge from-layer="652" from-port="2" to-layer="654" to-port="0" />
+		<edge from-layer="653" from-port="0" to-layer="654" to-port="1" />
+		<edge from-layer="654" from-port="2" to-layer="735" to-port="0" />
+		<edge from-layer="654" from-port="2" to-layer="696" to-port="0" />
+		<edge from-layer="654" from-port="2" to-layer="658" to-port="0" />
+		<edge from-layer="655" from-port="0" to-layer="658" to-port="1" />
+		<edge from-layer="656" from-port="0" to-layer="658" to-port="2" />
+		<edge from-layer="657" from-port="0" to-layer="658" to-port="3" />
+		<edge from-layer="658" from-port="5" to-layer="660" to-port="0" />
+		<edge from-layer="659" from-port="0" to-layer="660" to-port="1" />
+		<edge from-layer="660" from-port="2" to-layer="682" to-port="0" />
+		<edge from-layer="660" from-port="2" to-layer="664" to-port="0" />
+		<edge from-layer="661" from-port="0" to-layer="664" to-port="1" />
+		<edge from-layer="662" from-port="0" to-layer="664" to-port="2" />
+		<edge from-layer="663" from-port="0" to-layer="664" to-port="3" />
+		<edge from-layer="664" from-port="5" to-layer="665" to-port="0" />
+		<edge from-layer="664" from-port="5" to-layer="669" to-port="0" />
+		<edge from-layer="664" from-port="5" to-layer="675" to-port="0" />
+		<edge from-layer="665" from-port="2" to-layer="678" to-port="0" />
+		<edge from-layer="666" from-port="0" to-layer="669" to-port="1" />
+		<edge from-layer="667" from-port="0" to-layer="669" to-port="2" />
+		<edge from-layer="668" from-port="0" to-layer="669" to-port="3" />
+		<edge from-layer="669" from-port="5" to-layer="671" to-port="0" />
+		<edge from-layer="670" from-port="0" to-layer="671" to-port="1" />
+		<edge from-layer="671" from-port="2" to-layer="676" to-port="0" />
+		<edge from-layer="672" from-port="0" to-layer="675" to-port="1" />
+		<edge from-layer="673" from-port="0" to-layer="675" to-port="2" />
+		<edge from-layer="674" from-port="0" to-layer="675" to-port="3" />
+		<edge from-layer="675" from-port="5" to-layer="676" to-port="1" />
+		<edge from-layer="676" from-port="2" to-layer="677" to-port="0" />
+		<edge from-layer="677" from-port="2" to-layer="678" to-port="1" />
+		<edge from-layer="678" from-port="2" to-layer="683" to-port="0" />
+		<edge from-layer="679" from-port="0" to-layer="682" to-port="1" />
+		<edge from-layer="680" from-port="0" to-layer="682" to-port="2" />
+		<edge from-layer="681" from-port="0" to-layer="682" to-port="3" />
+		<edge from-layer="682" from-port="5" to-layer="683" to-port="1" />
+		<edge from-layer="683" from-port="2" to-layer="747" to-port="0" />
+		<edge from-layer="684" from-port="0" to-layer="689" to-port="0" />
+		<edge from-layer="685" from-port="0" to-layer="688" to-port="1" />
 		<edge from-layer="686" from-port="0" to-layer="688" to-port="2" />
 		<edge from-layer="687" from-port="0" to-layer="688" to-port="3" />
-		<edge from-layer="688" from-port="4" to-layer="691" to-port="0" />
-		<edge from-layer="689" from-port="0" to-layer="690" to-port="0" />
-		<edge from-layer="690" from-port="1" to-layer="691" to-port="1" />
-		<edge from-layer="691" from-port="2" to-layer="695" to-port="0" />
-		<edge from-layer="692" from-port="0" to-layer="694" to-port="1" />
-		<edge from-layer="693" from-port="0" to-layer="694" to-port="3" />
-		<edge from-layer="694" from-port="4" to-layer="695" to-port="1" />
-		<edge from-layer="695" from-port="2" to-layer="696" to-port="0" />
-		<edge from-layer="696" from-port="2" to-layer="697" to-port="1" />
-		<edge from-layer="697" from-port="2" to-layer="702" to-port="0" />
-		<edge from-layer="698" from-port="0" to-layer="701" to-port="1" />
-		<edge from-layer="699" from-port="0" to-layer="701" to-port="2" />
-		<edge from-layer="700" from-port="0" to-layer="701" to-port="3" />
-		<edge from-layer="701" from-port="4" to-layer="702" to-port="1" />
-		<edge from-layer="702" from-port="2" to-layer="703" to-port="1" />
-		<edge from-layer="703" from-port="2" to-layer="7698" to-port="0" />
-		<edge from-layer="703" from-port="2" to-layer="760" to-port="1" />
-		<edge from-layer="703" from-port="2" to-layer="724" to-port="0" />
-		<edge from-layer="704" from-port="0" to-layer="705" to-port="0" />
-		<edge from-layer="705" from-port="1" to-layer="710" to-port="0" />
-		<edge from-layer="706" from-port="0" to-layer="709" to-port="1" />
-		<edge from-layer="707" from-port="0" to-layer="709" to-port="2" />
-		<edge from-layer="708" from-port="0" to-layer="709" to-port="3" />
-		<edge from-layer="709" from-port="4" to-layer="710" to-port="1" />
-		<edge from-layer="710" from-port="2" to-layer="711" to-port="0" />
-		<edge from-layer="711" from-port="1" to-layer="713" to-port="0" />
-		<edge from-layer="712" from-port="0" to-layer="713" to-port="2" />
-		<edge from-layer="713" from-port="3" to-layer="720" to-port="0" />
-		<edge from-layer="714" from-port="0" to-layer="717" to-port="1" />
-		<edge from-layer="715" from-port="0" to-layer="717" to-port="2" />
-		<edge from-layer="716" from-port="0" to-layer="717" to-port="3" />
-		<edge from-layer="717" from-port="4" to-layer="719" to-port="0" />
-		<edge from-layer="718" from-port="0" to-layer="719" to-port="1" />
-		<edge from-layer="719" from-port="2" to-layer="720" to-port="1" />
-		<edge from-layer="720" from-port="2" to-layer="7697" to-port="0" />
-		<edge from-layer="720" from-port="2" to-layer="760" to-port="2" />
-		<edge from-layer="721" from-port="0" to-layer="739" to-port="0" />
-		<edge from-layer="722" from-port="0" to-layer="734" to-port="0" />
-		<edge from-layer="723" from-port="0" to-layer="737" to-port="1" />
-		<edge from-layer="723" from-port="0" to-layer="734" to-port="1" />
-		<edge from-layer="724" from-port="1" to-layer="726" to-port="0" />
-		<edge from-layer="725" from-port="0" to-layer="726" to-port="2" />
-		<edge from-layer="726" from-port="3" to-layer="730" to-port="0" />
-		<edge from-layer="726" from-port="3" to-layer="736" to-port="0" />
-		<edge from-layer="727" from-port="1" to-layer="755" to-port="0" />
-		<edge from-layer="727" from-port="1" to-layer="729" to-port="0" />
-		<edge from-layer="728" from-port="0" to-layer="729" to-port="2" />
-		<edge from-layer="729" from-port="3" to-layer="730" to-port="1" />
-		<edge from-layer="730" from-port="2" to-layer="732" to-port="0" />
-		<edge from-layer="731" from-port="0" to-layer="732" to-port="1" />
-		<edge from-layer="731" from-port="0" to-layer="736" to-port="1" />
-		<edge from-layer="732" from-port="2" to-layer="734" to-port="2" />
-		<edge from-layer="733" from-port="0" to-layer="737" to-port="3" />
-		<edge from-layer="733" from-port="0" to-layer="734" to-port="3" />
-		<edge from-layer="734" from-port="4" to-layer="739" to-port="1" />
-		<edge from-layer="735" from-port="0" to-layer="737" to-port="0" />
-		<edge from-layer="736" from-port="2" to-layer="737" to-port="2" />
-		<edge from-layer="736" from-port="2" to-layer="744" to-port="2" />
-		<edge from-layer="737" from-port="4" to-layer="739" to-port="2" />
-		<edge from-layer="738" from-port="0" to-layer="739" to-port="3" />
-		<edge from-layer="739" from-port="4" to-layer="746" to-port="0" />
-		<edge from-layer="740" from-port="0" to-layer="746" to-port="1" />
-		<edge from-layer="741" from-port="0" to-layer="744" to-port="0" />
-		<edge from-layer="742" from-port="0" to-layer="744" to-port="1" />
-		<edge from-layer="743" from-port="0" to-layer="744" to-port="3" />
-		<edge from-layer="744" from-port="4" to-layer="746" to-port="2" />
+		<edge from-layer="688" from-port="4" to-layer="689" to-port="1" />
+		<edge from-layer="689" from-port="2" to-layer="690" to-port="0" />
+		<edge from-layer="690" from-port="1" to-layer="692" to-port="0" />
+		<edge from-layer="691" from-port="0" to-layer="692" to-port="2" />
+		<edge from-layer="692" from-port="3" to-layer="740" to-port="0" />
+		<edge from-layer="692" from-port="3" to-layer="722" to-port="0" />
+		<edge from-layer="693" from-port="0" to-layer="696" to-port="1" />
+		<edge from-layer="694" from-port="0" to-layer="696" to-port="2" />
+		<edge from-layer="695" from-port="0" to-layer="696" to-port="3" />
+		<edge from-layer="696" from-port="5" to-layer="698" to-port="0" />
+		<edge from-layer="697" from-port="0" to-layer="698" to-port="1" />
+		<edge from-layer="698" from-port="2" to-layer="720" to-port="0" />
+		<edge from-layer="698" from-port="2" to-layer="702" to-port="0" />
+		<edge from-layer="699" from-port="0" to-layer="702" to-port="1" />
+		<edge from-layer="700" from-port="0" to-layer="702" to-port="2" />
+		<edge from-layer="701" from-port="0" to-layer="702" to-port="3" />
+		<edge from-layer="702" from-port="5" to-layer="707" to-port="0" />
+		<edge from-layer="702" from-port="5" to-layer="713" to-port="0" />
+		<edge from-layer="702" from-port="5" to-layer="703" to-port="0" />
+		<edge from-layer="703" from-port="2" to-layer="716" to-port="0" />
+		<edge from-layer="704" from-port="0" to-layer="707" to-port="1" />
+		<edge from-layer="705" from-port="0" to-layer="707" to-port="2" />
+		<edge from-layer="706" from-port="0" to-layer="707" to-port="3" />
+		<edge from-layer="707" from-port="5" to-layer="709" to-port="0" />
+		<edge from-layer="708" from-port="0" to-layer="709" to-port="1" />
+		<edge from-layer="709" from-port="2" to-layer="714" to-port="0" />
+		<edge from-layer="710" from-port="0" to-layer="713" to-port="1" />
+		<edge from-layer="711" from-port="0" to-layer="713" to-port="2" />
+		<edge from-layer="712" from-port="0" to-layer="713" to-port="3" />
+		<edge from-layer="713" from-port="5" to-layer="714" to-port="1" />
+		<edge from-layer="714" from-port="2" to-layer="715" to-port="0" />
+		<edge from-layer="715" from-port="2" to-layer="716" to-port="1" />
+		<edge from-layer="716" from-port="2" to-layer="721" to-port="0" />
+		<edge from-layer="717" from-port="0" to-layer="720" to-port="1" />
+		<edge from-layer="718" from-port="0" to-layer="720" to-port="2" />
+		<edge from-layer="719" from-port="0" to-layer="720" to-port="3" />
+		<edge from-layer="720" from-port="5" to-layer="721" to-port="1" />
+		<edge from-layer="721" from-port="2" to-layer="722" to-port="1" />
+		<edge from-layer="722" from-port="2" to-layer="5113" to-port="0" />
+		<edge from-layer="722" from-port="2" to-layer="747" to-port="1" />
+		<edge from-layer="723" from-port="0" to-layer="728" to-port="0" />
+		<edge from-layer="724" from-port="0" to-layer="727" to-port="1" />
+		<edge from-layer="725" from-port="0" to-layer="727" to-port="2" />
+		<edge from-layer="726" from-port="0" to-layer="727" to-port="3" />
+		<edge from-layer="727" from-port="4" to-layer="728" to-port="1" />
+		<edge from-layer="728" from-port="2" to-layer="729" to-port="0" />
+		<edge from-layer="729" from-port="1" to-layer="731" to-port="0" />
+		<edge from-layer="730" from-port="0" to-layer="731" to-port="2" />
+		<edge from-layer="731" from-port="3" to-layer="738" to-port="0" />
+		<edge from-layer="732" from-port="0" to-layer="735" to-port="1" />
+		<edge from-layer="733" from-port="0" to-layer="735" to-port="2" />
+		<edge from-layer="734" from-port="0" to-layer="735" to-port="3" />
+		<edge from-layer="735" from-port="5" to-layer="737" to-port="0" />
+		<edge from-layer="736" from-port="0" to-layer="737" to-port="1" />
+		<edge from-layer="737" from-port="2" to-layer="738" to-port="1" />
+		<edge from-layer="738" from-port="2" to-layer="747" to-port="2" />
+		<edge from-layer="738" from-port="2" to-layer="5112" to-port="0" />
+		<edge from-layer="739" from-port="0" to-layer="746" to-port="1" />
+		<edge from-layer="740" from-port="1" to-layer="743" to-port="0" />
+		<edge from-layer="741" from-port="0" to-layer="743" to-port="1" />
+		<edge from-layer="742" from-port="0" to-layer="743" to-port="2" />
+		<edge from-layer="743" from-port="3" to-layer="744" to-port="0" />
+		<edge from-layer="744" from-port="2" to-layer="746" to-port="2" />
 		<edge from-layer="745" from-port="0" to-layer="746" to-port="3" />
-		<edge from-layer="746" from-port="4" to-layer="750" to-port="0" />
-		<edge from-layer="747" from-port="0" to-layer="748" to-port="0" />
-		<edge from-layer="748" from-port="1" to-layer="750" to-port="1" />
-		<edge from-layer="749" from-port="0" to-layer="750" to-port="2" />
-		<edge from-layer="750" from-port="3" to-layer="751" to-port="0" />
-		<edge from-layer="751" from-port="2" to-layer="759" to-port="0" />
-		<edge from-layer="751" from-port="2" to-layer="752" to-port="0" />
-		<edge from-layer="752" from-port="1" to-layer="758" to-port="0" />
-		<edge from-layer="753" from-port="0" to-layer="755" to-port="1" />
-		<edge from-layer="754" from-port="0" to-layer="755" to-port="2" />
-		<edge from-layer="755" from-port="3" to-layer="756" to-port="0" />
-		<edge from-layer="756" from-port="4" to-layer="757" to-port="0" />
-		<edge from-layer="757" from-port="1" to-layer="758" to-port="1" />
-		<edge from-layer="758" from-port="2" to-layer="759" to-port="1" />
-		<edge from-layer="759" from-port="2" to-layer="760" to-port="3" />
-		<edge from-layer="760" from-port="4" to-layer="762" to-port="0" />
+		<edge from-layer="746" from-port="5" to-layer="747" to-port="3" />
+		<edge from-layer="747" from-port="4" to-layer="749" to-port="0" />
+		<edge from-layer="748" from-port="0" to-layer="749" to-port="1" />
+		<edge from-layer="749" from-port="2" to-layer="751" to-port="0" />
+		<edge from-layer="750" from-port="0" to-layer="751" to-port="1" />
+		<edge from-layer="751" from-port="2" to-layer="760" to-port="0" />
+		<edge from-layer="752" from-port="0" to-layer="753" to-port="0" />
+		<edge from-layer="753" from-port="1" to-layer="756" to-port="0" />
+		<edge from-layer="754" from-port="0" to-layer="755" to-port="0" />
+		<edge from-layer="755" from-port="1" to-layer="756" to-port="1" />
+		<edge from-layer="756" from-port="2" to-layer="758" to-port="0" />
+		<edge from-layer="757" from-port="0" to-layer="758" to-port="1" />
+		<edge from-layer="758" from-port="2" to-layer="759" to-port="0" />
+		<edge from-layer="759" from-port="1" to-layer="760" to-port="1" />
+		<edge from-layer="760" from-port="2" to-layer="762" to-port="0" />
 		<edge from-layer="761" from-port="0" to-layer="762" to-port="1" />
-		<edge from-layer="762" from-port="2" to-layer="764" to-port="0" />
-		<edge from-layer="763" from-port="0" to-layer="764" to-port="1" />
-		<edge from-layer="764" from-port="2" to-layer="773" to-port="0" />
-		<edge from-layer="765" from-port="0" to-layer="766" to-port="0" />
-		<edge from-layer="766" from-port="1" to-layer="769" to-port="0" />
-		<edge from-layer="767" from-port="0" to-layer="768" to-port="0" />
-		<edge from-layer="768" from-port="1" to-layer="769" to-port="1" />
-		<edge from-layer="769" from-port="2" to-layer="771" to-port="0" />
-		<edge from-layer="770" from-port="0" to-layer="771" to-port="1" />
-		<edge from-layer="771" from-port="2" to-layer="772" to-port="0" />
-		<edge from-layer="772" from-port="1" to-layer="773" to-port="1" />
-		<edge from-layer="773" from-port="2" to-layer="776" to-port="0" />
-		<edge from-layer="774" from-port="0" to-layer="775" to-port="0" />
-		<edge from-layer="775" from-port="1" to-layer="776" to-port="1" />
-		<edge from-layer="776" from-port="2" to-layer="777" to-port="1" />
+		<edge from-layer="762" from-port="2" to-layer="763" to-port="1" />
+		<edge from-layer="763" from-port="2" to-layer="764" to-port="0" />
+		<edge from-layer="764" from-port="2" to-layer="766" to-port="0" />
+		<edge from-layer="764" from-port="2" to-layer="922" to-port="1" />
+		<edge from-layer="765" from-port="0" to-layer="766" to-port="1" />
+		<edge from-layer="766" from-port="2" to-layer="797" to-port="0" />
+		<edge from-layer="766" from-port="2" to-layer="768" to-port="0" />
+		<edge from-layer="767" from-port="0" to-layer="768" to-port="1" />
+		<edge from-layer="768" from-port="2" to-layer="770" to-port="0" />
+		<edge from-layer="769" from-port="0" to-layer="770" to-port="1" />
+		<edge from-layer="770" from-port="2" to-layer="779" to-port="0" />
+		<edge from-layer="771" from-port="0" to-layer="772" to-port="0" />
+		<edge from-layer="772" from-port="1" to-layer="775" to-port="0" />
+		<edge from-layer="773" from-port="0" to-layer="774" to-port="0" />
+		<edge from-layer="774" from-port="1" to-layer="775" to-port="1" />
+		<edge from-layer="775" from-port="2" to-layer="777" to-port="0" />
+		<edge from-layer="776" from-port="0" to-layer="777" to-port="1" />
 		<edge from-layer="777" from-port="2" to-layer="778" to-port="0" />
-		<edge from-layer="778" from-port="2" to-layer="1020" to-port="1" />
-		<edge from-layer="778" from-port="2" to-layer="780" to-port="0" />
-		<edge from-layer="779" from-port="0" to-layer="780" to-port="1" />
-		<edge from-layer="780" from-port="2" to-layer="814" to-port="0" />
-		<edge from-layer="780" from-port="2" to-layer="783" to-port="0" />
-		<edge from-layer="781" from-port="0" to-layer="782" to-port="0" />
-		<edge from-layer="782" from-port="1" to-layer="783" to-port="1" />
-		<edge from-layer="783" from-port="2" to-layer="786" to-port="0" />
+		<edge from-layer="778" from-port="1" to-layer="779" to-port="1" />
+		<edge from-layer="779" from-port="2" to-layer="782" to-port="0" />
+		<edge from-layer="780" from-port="0" to-layer="781" to-port="0" />
+		<edge from-layer="781" from-port="1" to-layer="782" to-port="1" />
+		<edge from-layer="782" from-port="2" to-layer="783" to-port="0" />
+		<edge from-layer="783" from-port="1" to-layer="792" to-port="0" />
 		<edge from-layer="784" from-port="0" to-layer="785" to-port="0" />
-		<edge from-layer="785" from-port="1" to-layer="786" to-port="1" />
-		<edge from-layer="786" from-port="2" to-layer="795" to-port="0" />
-		<edge from-layer="787" from-port="0" to-layer="788" to-port="0" />
-		<edge from-layer="788" from-port="1" to-layer="791" to-port="0" />
-		<edge from-layer="789" from-port="0" to-layer="790" to-port="0" />
-		<edge from-layer="790" from-port="1" to-layer="791" to-port="1" />
-		<edge from-layer="791" from-port="2" to-layer="793" to-port="0" />
-		<edge from-layer="792" from-port="0" to-layer="793" to-port="1" />
-		<edge from-layer="793" from-port="2" to-layer="794" to-port="0" />
+		<edge from-layer="785" from-port="1" to-layer="788" to-port="0" />
+		<edge from-layer="786" from-port="0" to-layer="787" to-port="0" />
+		<edge from-layer="787" from-port="1" to-layer="788" to-port="1" />
+		<edge from-layer="788" from-port="2" to-layer="790" to-port="0" />
+		<edge from-layer="789" from-port="0" to-layer="790" to-port="1" />
+		<edge from-layer="790" from-port="2" to-layer="791" to-port="0" />
+		<edge from-layer="791" from-port="1" to-layer="792" to-port="1" />
+		<edge from-layer="792" from-port="2" to-layer="795" to-port="0" />
+		<edge from-layer="793" from-port="0" to-layer="794" to-port="0" />
 		<edge from-layer="794" from-port="1" to-layer="795" to-port="1" />
-		<edge from-layer="795" from-port="2" to-layer="798" to-port="0" />
-		<edge from-layer="796" from-port="0" to-layer="797" to-port="0" />
-		<edge from-layer="797" from-port="1" to-layer="798" to-port="1" />
-		<edge from-layer="798" from-port="2" to-layer="799" to-port="0" />
-		<edge from-layer="799" from-port="1" to-layer="808" to-port="0" />
+		<edge from-layer="795" from-port="2" to-layer="921" to-port="0" />
+		<edge from-layer="796" from-port="0" to-layer="797" to-port="1" />
+		<edge from-layer="797" from-port="2" to-layer="799" to-port="0" />
+		<edge from-layer="798" from-port="0" to-layer="799" to-port="1" />
+		<edge from-layer="799" from-port="2" to-layer="808" to-port="0" />
 		<edge from-layer="800" from-port="0" to-layer="801" to-port="0" />
 		<edge from-layer="801" from-port="1" to-layer="804" to-port="0" />
 		<edge from-layer="802" from-port="0" to-layer="803" to-port="0" />
@@ -122438,1375 +82070,1315 @@
 		<edge from-layer="805" from-port="0" to-layer="806" to-port="1" />
 		<edge from-layer="806" from-port="2" to-layer="807" to-port="0" />
 		<edge from-layer="807" from-port="1" to-layer="808" to-port="1" />
-		<edge from-layer="808" from-port="2" to-layer="811" to-port="0" />
-		<edge from-layer="809" from-port="0" to-layer="810" to-port="0" />
-		<edge from-layer="810" from-port="1" to-layer="811" to-port="1" />
-		<edge from-layer="811" from-port="2" to-layer="1019" to-port="0" />
-		<edge from-layer="812" from-port="0" to-layer="813" to-port="0" />
-		<edge from-layer="813" from-port="1" to-layer="814" to-port="1" />
-		<edge from-layer="814" from-port="2" to-layer="817" to-port="0" />
-		<edge from-layer="815" from-port="0" to-layer="816" to-port="0" />
-		<edge from-layer="816" from-port="1" to-layer="817" to-port="1" />
-		<edge from-layer="817" from-port="2" to-layer="826" to-port="0" />
-		<edge from-layer="818" from-port="0" to-layer="819" to-port="0" />
-		<edge from-layer="819" from-port="1" to-layer="822" to-port="0" />
-		<edge from-layer="820" from-port="0" to-layer="821" to-port="0" />
-		<edge from-layer="821" from-port="1" to-layer="822" to-port="1" />
-		<edge from-layer="822" from-port="2" to-layer="824" to-port="0" />
-		<edge from-layer="823" from-port="0" to-layer="824" to-port="1" />
-		<edge from-layer="824" from-port="2" to-layer="825" to-port="0" />
-		<edge from-layer="825" from-port="1" to-layer="826" to-port="1" />
-		<edge from-layer="826" from-port="2" to-layer="829" to-port="0" />
-		<edge from-layer="827" from-port="0" to-layer="828" to-port="0" />
-		<edge from-layer="828" from-port="1" to-layer="829" to-port="1" />
-		<edge from-layer="829" from-port="2" to-layer="831" to-port="0" />
-		<edge from-layer="830" from-port="0" to-layer="831" to-port="1" />
-		<edge from-layer="831" from-port="2" to-layer="959" to-port="0" />
-		<edge from-layer="831" from-port="2" to-layer="835" to-port="0" />
-		<edge from-layer="831" from-port="2" to-layer="848" to-port="0" />
-		<edge from-layer="832" from-port="0" to-layer="835" to-port="1" />
-		<edge from-layer="833" from-port="0" to-layer="835" to-port="2" />
-		<edge from-layer="834" from-port="0" to-layer="835" to-port="3" />
-		<edge from-layer="835" from-port="4" to-layer="837" to-port="0" />
-		<edge from-layer="836" from-port="0" to-layer="837" to-port="1" />
-		<edge from-layer="837" from-port="2" to-layer="841" to-port="0" />
-		<edge from-layer="837" from-port="2" to-layer="910" to-port="0" />
-		<edge from-layer="838" from-port="0" to-layer="841" to-port="1" />
-		<edge from-layer="839" from-port="0" to-layer="841" to-port="2" />
-		<edge from-layer="840" from-port="0" to-layer="841" to-port="3" />
-		<edge from-layer="841" from-port="4" to-layer="875" to-port="0" />
-		<edge from-layer="841" from-port="4" to-layer="895" to-port="0" />
-		<edge from-layer="841" from-port="4" to-layer="878" to-port="0" />
-		<edge from-layer="841" from-port="4" to-layer="889" to-port="0" />
-		<edge from-layer="842" from-port="0" to-layer="843" to-port="0" />
-		<edge from-layer="843" from-port="1" to-layer="871" to-port="0" />
-		<edge from-layer="844" from-port="0" to-layer="871" to-port="1" />
-		<edge from-layer="845" from-port="0" to-layer="848" to-port="1" />
-		<edge from-layer="846" from-port="0" to-layer="848" to-port="2" />
-		<edge from-layer="847" from-port="0" to-layer="848" to-port="3" />
-		<edge from-layer="848" from-port="4" to-layer="850" to-port="0" />
-		<edge from-layer="849" from-port="0" to-layer="850" to-port="1" />
-		<edge from-layer="850" from-port="2" to-layer="851" to-port="0" />
-		<edge from-layer="850" from-port="2" to-layer="943" to-port="0" />
-		<edge from-layer="850" from-port="2" to-layer="915" to-port="0" />
-		<edge from-layer="851" from-port="1" to-layer="853" to-port="0" />
-		<edge from-layer="852" from-port="0" to-layer="853" to-port="2" />
-		<edge from-layer="853" from-port="3" to-layer="867" to-port="0" />
-		<edge from-layer="854" from-port="0" to-layer="855" to-port="0" />
-		<edge from-layer="855" from-port="1" to-layer="860" to-port="0" />
-		<edge from-layer="856" from-port="0" to-layer="859" to-port="1" />
-		<edge from-layer="857" from-port="0" to-layer="859" to-port="2" />
-		<edge from-layer="858" from-port="0" to-layer="859" to-port="3" />
-		<edge from-layer="859" from-port="4" to-layer="860" to-port="1" />
-		<edge from-layer="860" from-port="2" to-layer="861" to-port="0" />
-		<edge from-layer="861" from-port="1" to-layer="863" to-port="0" />
-		<edge from-layer="862" from-port="0" to-layer="863" to-port="2" />
-		<edge from-layer="863" from-port="3" to-layer="945" to-port="0" />
-		<edge from-layer="863" from-port="3" to-layer="864" to-port="0" />
-		<edge from-layer="864" from-port="1" to-layer="866" to-port="0" />
-		<edge from-layer="865" from-port="0" to-layer="866" to-port="2" />
-		<edge from-layer="866" from-port="3" to-layer="867" to-port="1" />
-		<edge from-layer="867" from-port="2" to-layer="869" to-port="0" />
-		<edge from-layer="868" from-port="0" to-layer="869" to-port="1" />
-		<edge from-layer="869" from-port="2" to-layer="901" to-port="2" />
-		<edge from-layer="869" from-port="2" to-layer="871" to-port="2" />
+		<edge from-layer="808" from-port="2" to-layer="810" to-port="0" />
+		<edge from-layer="809" from-port="0" to-layer="810" to-port="1" />
+		<edge from-layer="810" from-port="2" to-layer="812" to-port="0" />
+		<edge from-layer="811" from-port="0" to-layer="812" to-port="1" />
+		<edge from-layer="812" from-port="2" to-layer="893" to-port="0" />
+		<edge from-layer="812" from-port="2" to-layer="816" to-port="0" />
+		<edge from-layer="812" from-port="2" to-layer="854" to-port="0" />
+		<edge from-layer="813" from-port="0" to-layer="816" to-port="1" />
+		<edge from-layer="814" from-port="0" to-layer="816" to-port="2" />
+		<edge from-layer="815" from-port="0" to-layer="816" to-port="3" />
+		<edge from-layer="816" from-port="5" to-layer="818" to-port="0" />
+		<edge from-layer="817" from-port="0" to-layer="818" to-port="1" />
+		<edge from-layer="818" from-port="2" to-layer="840" to-port="0" />
+		<edge from-layer="818" from-port="2" to-layer="822" to-port="0" />
+		<edge from-layer="819" from-port="0" to-layer="822" to-port="1" />
+		<edge from-layer="820" from-port="0" to-layer="822" to-port="2" />
+		<edge from-layer="821" from-port="0" to-layer="822" to-port="3" />
+		<edge from-layer="822" from-port="5" to-layer="823" to-port="0" />
+		<edge from-layer="822" from-port="5" to-layer="833" to-port="0" />
+		<edge from-layer="822" from-port="5" to-layer="827" to-port="0" />
+		<edge from-layer="823" from-port="2" to-layer="836" to-port="0" />
+		<edge from-layer="824" from-port="0" to-layer="827" to-port="1" />
+		<edge from-layer="825" from-port="0" to-layer="827" to-port="2" />
+		<edge from-layer="826" from-port="0" to-layer="827" to-port="3" />
+		<edge from-layer="827" from-port="5" to-layer="829" to-port="0" />
+		<edge from-layer="828" from-port="0" to-layer="829" to-port="1" />
+		<edge from-layer="829" from-port="2" to-layer="834" to-port="0" />
+		<edge from-layer="830" from-port="0" to-layer="833" to-port="1" />
+		<edge from-layer="831" from-port="0" to-layer="833" to-port="2" />
+		<edge from-layer="832" from-port="0" to-layer="833" to-port="3" />
+		<edge from-layer="833" from-port="5" to-layer="834" to-port="1" />
+		<edge from-layer="834" from-port="2" to-layer="835" to-port="0" />
+		<edge from-layer="835" from-port="2" to-layer="836" to-port="1" />
+		<edge from-layer="836" from-port="2" to-layer="841" to-port="0" />
+		<edge from-layer="837" from-port="0" to-layer="840" to-port="1" />
+		<edge from-layer="838" from-port="0" to-layer="840" to-port="2" />
+		<edge from-layer="839" from-port="0" to-layer="840" to-port="3" />
+		<edge from-layer="840" from-port="5" to-layer="841" to-port="1" />
+		<edge from-layer="841" from-port="2" to-layer="905" to-port="0" />
+		<edge from-layer="842" from-port="0" to-layer="847" to-port="0" />
+		<edge from-layer="843" from-port="0" to-layer="846" to-port="1" />
+		<edge from-layer="844" from-port="0" to-layer="846" to-port="2" />
+		<edge from-layer="845" from-port="0" to-layer="846" to-port="3" />
+		<edge from-layer="846" from-port="4" to-layer="847" to-port="1" />
+		<edge from-layer="847" from-port="2" to-layer="848" to-port="0" />
+		<edge from-layer="848" from-port="1" to-layer="850" to-port="0" />
+		<edge from-layer="849" from-port="0" to-layer="850" to-port="2" />
+		<edge from-layer="850" from-port="3" to-layer="898" to-port="0" />
+		<edge from-layer="850" from-port="3" to-layer="880" to-port="0" />
+		<edge from-layer="851" from-port="0" to-layer="854" to-port="1" />
+		<edge from-layer="852" from-port="0" to-layer="854" to-port="2" />
+		<edge from-layer="853" from-port="0" to-layer="854" to-port="3" />
+		<edge from-layer="854" from-port="5" to-layer="856" to-port="0" />
+		<edge from-layer="855" from-port="0" to-layer="856" to-port="1" />
+		<edge from-layer="856" from-port="2" to-layer="860" to-port="0" />
+		<edge from-layer="856" from-port="2" to-layer="878" to-port="0" />
+		<edge from-layer="857" from-port="0" to-layer="860" to-port="1" />
+		<edge from-layer="858" from-port="0" to-layer="860" to-port="2" />
+		<edge from-layer="859" from-port="0" to-layer="860" to-port="3" />
+		<edge from-layer="860" from-port="5" to-layer="871" to-port="0" />
+		<edge from-layer="860" from-port="5" to-layer="865" to-port="0" />
+		<edge from-layer="860" from-port="5" to-layer="861" to-port="0" />
+		<edge from-layer="861" from-port="2" to-layer="874" to-port="0" />
+		<edge from-layer="862" from-port="0" to-layer="865" to-port="1" />
+		<edge from-layer="863" from-port="0" to-layer="865" to-port="2" />
+		<edge from-layer="864" from-port="0" to-layer="865" to-port="3" />
+		<edge from-layer="865" from-port="5" to-layer="867" to-port="0" />
+		<edge from-layer="866" from-port="0" to-layer="867" to-port="1" />
+		<edge from-layer="867" from-port="2" to-layer="872" to-port="0" />
+		<edge from-layer="868" from-port="0" to-layer="871" to-port="1" />
+		<edge from-layer="869" from-port="0" to-layer="871" to-port="2" />
 		<edge from-layer="870" from-port="0" to-layer="871" to-port="3" />
-		<edge from-layer="871" from-port="4" to-layer="873" to-port="0" />
-		<edge from-layer="872" from-port="0" to-layer="873" to-port="2" />
-		<edge from-layer="873" from-port="3" to-layer="874" to-port="0" />
-		<edge from-layer="874" from-port="2" to-layer="916" to-port="1" />
-		<edge from-layer="874" from-port="2" to-layer="875" to-port="1" />
-		<edge from-layer="875" from-port="2" to-layer="906" to-port="0" />
-		<edge from-layer="876" from-port="0" to-layer="886" to-port="0" />
-		<edge from-layer="877" from-port="0" to-layer="886" to-port="1" />
-		<edge from-layer="878" from-port="1" to-layer="880" to-port="0" />
-		<edge from-layer="879" from-port="0" to-layer="880" to-port="2" />
-		<edge from-layer="880" from-port="3" to-layer="881" to-port="0" />
-		<edge from-layer="881" from-port="2" to-layer="882" to-port="0" />
-		<edge from-layer="882" from-port="1" to-layer="884" to-port="0" />
-		<edge from-layer="883" from-port="0" to-layer="884" to-port="1" />
-		<edge from-layer="884" from-port="2" to-layer="886" to-port="2" />
-		<edge from-layer="885" from-port="0" to-layer="886" to-port="3" />
-		<edge from-layer="886" from-port="4" to-layer="889" to-port="1" />
-		<edge from-layer="886" from-port="4" to-layer="895" to-port="2" />
-		<edge from-layer="887" from-port="0" to-layer="889" to-port="2" />
-		<edge from-layer="888" from-port="0" to-layer="889" to-port="3" />
-		<edge from-layer="889" from-port="4" to-layer="892" to-port="0" />
-		<edge from-layer="890" from-port="0" to-layer="891" to-port="0" />
-		<edge from-layer="891" from-port="1" to-layer="892" to-port="1" />
-		<edge from-layer="892" from-port="2" to-layer="896" to-port="0" />
-		<edge from-layer="893" from-port="0" to-layer="895" to-port="1" />
-		<edge from-layer="894" from-port="0" to-layer="895" to-port="3" />
-		<edge from-layer="895" from-port="4" to-layer="896" to-port="1" />
-		<edge from-layer="896" from-port="2" to-layer="905" to-port="0" />
-		<edge from-layer="897" from-port="0" to-layer="898" to-port="0" />
+		<edge from-layer="871" from-port="5" to-layer="872" to-port="1" />
+		<edge from-layer="872" from-port="2" to-layer="873" to-port="0" />
+		<edge from-layer="873" from-port="2" to-layer="874" to-port="1" />
+		<edge from-layer="874" from-port="2" to-layer="879" to-port="0" />
+		<edge from-layer="875" from-port="0" to-layer="878" to-port="1" />
+		<edge from-layer="876" from-port="0" to-layer="878" to-port="2" />
+		<edge from-layer="877" from-port="0" to-layer="878" to-port="3" />
+		<edge from-layer="878" from-port="5" to-layer="879" to-port="1" />
+		<edge from-layer="879" from-port="2" to-layer="880" to-port="1" />
+		<edge from-layer="880" from-port="2" to-layer="905" to-port="1" />
+		<edge from-layer="880" from-port="2" to-layer="1631" to-port="0" />
+		<edge from-layer="881" from-port="0" to-layer="886" to-port="0" />
+		<edge from-layer="882" from-port="0" to-layer="885" to-port="1" />
+		<edge from-layer="883" from-port="0" to-layer="885" to-port="2" />
+		<edge from-layer="884" from-port="0" to-layer="885" to-port="3" />
+		<edge from-layer="885" from-port="4" to-layer="886" to-port="1" />
+		<edge from-layer="886" from-port="2" to-layer="887" to-port="0" />
+		<edge from-layer="887" from-port="1" to-layer="889" to-port="0" />
+		<edge from-layer="888" from-port="0" to-layer="889" to-port="2" />
+		<edge from-layer="889" from-port="3" to-layer="896" to-port="0" />
+		<edge from-layer="890" from-port="0" to-layer="893" to-port="1" />
+		<edge from-layer="891" from-port="0" to-layer="893" to-port="2" />
+		<edge from-layer="892" from-port="0" to-layer="893" to-port="3" />
+		<edge from-layer="893" from-port="5" to-layer="895" to-port="0" />
+		<edge from-layer="894" from-port="0" to-layer="895" to-port="1" />
+		<edge from-layer="895" from-port="2" to-layer="896" to-port="1" />
+		<edge from-layer="896" from-port="2" to-layer="1630" to-port="0" />
+		<edge from-layer="896" from-port="2" to-layer="905" to-port="2" />
+		<edge from-layer="897" from-port="0" to-layer="904" to-port="1" />
 		<edge from-layer="898" from-port="1" to-layer="901" to-port="0" />
 		<edge from-layer="899" from-port="0" to-layer="901" to-port="1" />
-		<edge from-layer="900" from-port="0" to-layer="901" to-port="3" />
-		<edge from-layer="901" from-port="4" to-layer="903" to-port="0" />
-		<edge from-layer="902" from-port="0" to-layer="903" to-port="2" />
-		<edge from-layer="903" from-port="3" to-layer="904" to-port="0" />
-		<edge from-layer="904" from-port="2" to-layer="938" to-port="1" />
-		<edge from-layer="904" from-port="2" to-layer="905" to-port="1" />
-		<edge from-layer="905" from-port="2" to-layer="906" to-port="1" />
-		<edge from-layer="906" from-port="2" to-layer="911" to-port="0" />
-		<edge from-layer="907" from-port="0" to-layer="910" to-port="1" />
-		<edge from-layer="908" from-port="0" to-layer="910" to-port="2" />
-		<edge from-layer="909" from-port="0" to-layer="910" to-port="3" />
-		<edge from-layer="910" from-port="4" to-layer="911" to-port="1" />
-		<edge from-layer="911" from-port="2" to-layer="969" to-port="0" />
-		<edge from-layer="911" from-port="2" to-layer="1002" to-port="0" />
-		<edge from-layer="912" from-port="0" to-layer="915" to-port="1" />
-		<edge from-layer="913" from-port="0" to-layer="915" to-port="2" />
-		<edge from-layer="914" from-port="0" to-layer="915" to-port="3" />
-		<edge from-layer="915" from-port="4" to-layer="936" to-port="0" />
-		<edge from-layer="915" from-port="4" to-layer="930" to-port="0" />
-		<edge from-layer="915" from-port="4" to-layer="919" to-port="0" />
-		<edge from-layer="915" from-port="4" to-layer="916" to-port="0" />
-		<edge from-layer="916" from-port="2" to-layer="939" to-port="0" />
-		<edge from-layer="917" from-port="0" to-layer="927" to-port="0" />
-		<edge from-layer="918" from-port="0" to-layer="927" to-port="1" />
-		<edge from-layer="919" from-port="1" to-layer="921" to-port="0" />
-		<edge from-layer="920" from-port="0" to-layer="921" to-port="2" />
-		<edge from-layer="921" from-port="3" to-layer="922" to-port="0" />
-		<edge from-layer="922" from-port="2" to-layer="923" to-port="0" />
-		<edge from-layer="923" from-port="1" to-layer="925" to-port="0" />
-		<edge from-layer="924" from-port="0" to-layer="925" to-port="1" />
-		<edge from-layer="925" from-port="2" to-layer="927" to-port="2" />
-		<edge from-layer="926" from-port="0" to-layer="927" to-port="3" />
-		<edge from-layer="927" from-port="4" to-layer="936" to-port="2" />
-		<edge from-layer="927" from-port="4" to-layer="930" to-port="1" />
-		<edge from-layer="928" from-port="0" to-layer="930" to-port="2" />
-		<edge from-layer="929" from-port="0" to-layer="930" to-port="3" />
-		<edge from-layer="930" from-port="4" to-layer="933" to-port="0" />
+		<edge from-layer="900" from-port="0" to-layer="901" to-port="2" />
+		<edge from-layer="901" from-port="3" to-layer="902" to-port="0" />
+		<edge from-layer="902" from-port="2" to-layer="904" to-port="2" />
+		<edge from-layer="903" from-port="0" to-layer="904" to-port="3" />
+		<edge from-layer="904" from-port="5" to-layer="905" to-port="3" />
+		<edge from-layer="905" from-port="4" to-layer="907" to-port="0" />
+		<edge from-layer="906" from-port="0" to-layer="907" to-port="1" />
+		<edge from-layer="907" from-port="2" to-layer="909" to-port="0" />
+		<edge from-layer="908" from-port="0" to-layer="909" to-port="1" />
+		<edge from-layer="909" from-port="2" to-layer="918" to-port="0" />
+		<edge from-layer="910" from-port="0" to-layer="911" to-port="0" />
+		<edge from-layer="911" from-port="1" to-layer="914" to-port="0" />
+		<edge from-layer="912" from-port="0" to-layer="913" to-port="0" />
+		<edge from-layer="913" from-port="1" to-layer="914" to-port="1" />
+		<edge from-layer="914" from-port="2" to-layer="916" to-port="0" />
+		<edge from-layer="915" from-port="0" to-layer="916" to-port="1" />
+		<edge from-layer="916" from-port="2" to-layer="917" to-port="0" />
+		<edge from-layer="917" from-port="1" to-layer="918" to-port="1" />
+		<edge from-layer="918" from-port="2" to-layer="920" to-port="0" />
+		<edge from-layer="919" from-port="0" to-layer="920" to-port="1" />
+		<edge from-layer="920" from-port="2" to-layer="921" to-port="1" />
+		<edge from-layer="921" from-port="2" to-layer="922" to-port="0" />
+		<edge from-layer="922" from-port="2" to-layer="1080" to-port="1" />
+		<edge from-layer="922" from-port="2" to-layer="924" to-port="0" />
+		<edge from-layer="923" from-port="0" to-layer="924" to-port="1" />
+		<edge from-layer="924" from-port="2" to-layer="955" to-port="0" />
+		<edge from-layer="924" from-port="2" to-layer="926" to-port="0" />
+		<edge from-layer="925" from-port="0" to-layer="926" to-port="1" />
+		<edge from-layer="926" from-port="2" to-layer="928" to-port="0" />
+		<edge from-layer="927" from-port="0" to-layer="928" to-port="1" />
+		<edge from-layer="928" from-port="2" to-layer="937" to-port="0" />
+		<edge from-layer="929" from-port="0" to-layer="930" to-port="0" />
+		<edge from-layer="930" from-port="1" to-layer="933" to-port="0" />
 		<edge from-layer="931" from-port="0" to-layer="932" to-port="0" />
 		<edge from-layer="932" from-port="1" to-layer="933" to-port="1" />
-		<edge from-layer="933" from-port="2" to-layer="937" to-port="0" />
-		<edge from-layer="934" from-port="0" to-layer="936" to-port="1" />
-		<edge from-layer="935" from-port="0" to-layer="936" to-port="3" />
-		<edge from-layer="936" from-port="4" to-layer="937" to-port="1" />
-		<edge from-layer="937" from-port="2" to-layer="938" to-port="0" />
-		<edge from-layer="938" from-port="2" to-layer="939" to-port="1" />
-		<edge from-layer="939" from-port="2" to-layer="944" to-port="0" />
-		<edge from-layer="940" from-port="0" to-layer="943" to-port="1" />
-		<edge from-layer="941" from-port="0" to-layer="943" to-port="2" />
-		<edge from-layer="942" from-port="0" to-layer="943" to-port="3" />
-		<edge from-layer="943" from-port="4" to-layer="944" to-port="1" />
-		<edge from-layer="944" from-port="2" to-layer="945" to-port="1" />
-		<edge from-layer="945" from-port="2" to-layer="7676" to-port="0" />
-		<edge from-layer="945" from-port="2" to-layer="966" to-port="0" />
-		<edge from-layer="945" from-port="2" to-layer="1002" to-port="1" />
-		<edge from-layer="946" from-port="0" to-layer="947" to-port="0" />
-		<edge from-layer="947" from-port="1" to-layer="952" to-port="0" />
-		<edge from-layer="948" from-port="0" to-layer="951" to-port="1" />
-		<edge from-layer="949" from-port="0" to-layer="951" to-port="2" />
-		<edge from-layer="950" from-port="0" to-layer="951" to-port="3" />
-		<edge from-layer="951" from-port="4" to-layer="952" to-port="1" />
-		<edge from-layer="952" from-port="2" to-layer="953" to-port="0" />
-		<edge from-layer="953" from-port="1" to-layer="955" to-port="0" />
-		<edge from-layer="954" from-port="0" to-layer="955" to-port="2" />
-		<edge from-layer="955" from-port="3" to-layer="962" to-port="0" />
-		<edge from-layer="956" from-port="0" to-layer="959" to-port="1" />
-		<edge from-layer="957" from-port="0" to-layer="959" to-port="2" />
-		<edge from-layer="958" from-port="0" to-layer="959" to-port="3" />
-		<edge from-layer="959" from-port="4" to-layer="961" to-port="0" />
-		<edge from-layer="960" from-port="0" to-layer="961" to-port="1" />
-		<edge from-layer="961" from-port="2" to-layer="962" to-port="1" />
-		<edge from-layer="962" from-port="2" to-layer="1002" to-port="2" />
-		<edge from-layer="962" from-port="2" to-layer="7675" to-port="0" />
-		<edge from-layer="963" from-port="0" to-layer="981" to-port="0" />
-		<edge from-layer="964" from-port="0" to-layer="976" to-port="0" />
-		<edge from-layer="965" from-port="0" to-layer="976" to-port="1" />
-		<edge from-layer="965" from-port="0" to-layer="979" to-port="1" />
-		<edge from-layer="966" from-port="1" to-layer="968" to-port="0" />
-		<edge from-layer="967" from-port="0" to-layer="968" to-port="2" />
-		<edge from-layer="968" from-port="3" to-layer="972" to-port="0" />
-		<edge from-layer="968" from-port="3" to-layer="978" to-port="0" />
-		<edge from-layer="969" from-port="1" to-layer="997" to-port="0" />
-		<edge from-layer="969" from-port="1" to-layer="971" to-port="0" />
-		<edge from-layer="970" from-port="0" to-layer="971" to-port="2" />
-		<edge from-layer="971" from-port="3" to-layer="972" to-port="1" />
-		<edge from-layer="972" from-port="2" to-layer="974" to-port="0" />
-		<edge from-layer="973" from-port="0" to-layer="978" to-port="1" />
-		<edge from-layer="973" from-port="0" to-layer="974" to-port="1" />
-		<edge from-layer="974" from-port="2" to-layer="976" to-port="2" />
-		<edge from-layer="975" from-port="0" to-layer="979" to-port="3" />
-		<edge from-layer="975" from-port="0" to-layer="976" to-port="3" />
-		<edge from-layer="976" from-port="4" to-layer="981" to-port="1" />
-		<edge from-layer="977" from-port="0" to-layer="979" to-port="0" />
-		<edge from-layer="978" from-port="2" to-layer="986" to-port="2" />
-		<edge from-layer="978" from-port="2" to-layer="979" to-port="2" />
-		<edge from-layer="979" from-port="4" to-layer="981" to-port="2" />
-		<edge from-layer="980" from-port="0" to-layer="981" to-port="3" />
-		<edge from-layer="981" from-port="4" to-layer="988" to-port="0" />
-		<edge from-layer="982" from-port="0" to-layer="988" to-port="1" />
-		<edge from-layer="983" from-port="0" to-layer="986" to-port="0" />
-		<edge from-layer="984" from-port="0" to-layer="986" to-port="1" />
-		<edge from-layer="985" from-port="0" to-layer="986" to-port="3" />
-		<edge from-layer="986" from-port="4" to-layer="988" to-port="2" />
-		<edge from-layer="987" from-port="0" to-layer="988" to-port="3" />
-		<edge from-layer="988" from-port="4" to-layer="992" to-port="0" />
-		<edge from-layer="989" from-port="0" to-layer="990" to-port="0" />
-		<edge from-layer="990" from-port="1" to-layer="992" to-port="1" />
-		<edge from-layer="991" from-port="0" to-layer="992" to-port="2" />
-		<edge from-layer="992" from-port="3" to-layer="993" to-port="0" />
-		<edge from-layer="993" from-port="2" to-layer="994" to-port="0" />
-		<edge from-layer="993" from-port="2" to-layer="1001" to-port="0" />
-		<edge from-layer="994" from-port="1" to-layer="1000" to-port="0" />
-		<edge from-layer="995" from-port="0" to-layer="997" to-port="1" />
-		<edge from-layer="996" from-port="0" to-layer="997" to-port="2" />
-		<edge from-layer="997" from-port="3" to-layer="998" to-port="0" />
-		<edge from-layer="998" from-port="4" to-layer="999" to-port="0" />
-		<edge from-layer="999" from-port="1" to-layer="1000" to-port="1" />
-		<edge from-layer="1000" from-port="2" to-layer="1001" to-port="1" />
-		<edge from-layer="1001" from-port="2" to-layer="1002" to-port="3" />
-		<edge from-layer="1002" from-port="4" to-layer="1004" to-port="0" />
-		<edge from-layer="1003" from-port="0" to-layer="1004" to-port="1" />
-		<edge from-layer="1004" from-port="2" to-layer="1006" to-port="0" />
-		<edge from-layer="1005" from-port="0" to-layer="1006" to-port="1" />
-		<edge from-layer="1006" from-port="2" to-layer="1015" to-port="0" />
-		<edge from-layer="1007" from-port="0" to-layer="1008" to-port="0" />
-		<edge from-layer="1008" from-port="1" to-layer="1011" to-port="0" />
-		<edge from-layer="1009" from-port="0" to-layer="1010" to-port="0" />
-		<edge from-layer="1010" from-port="1" to-layer="1011" to-port="1" />
-		<edge from-layer="1011" from-port="2" to-layer="1013" to-port="0" />
-		<edge from-layer="1012" from-port="0" to-layer="1013" to-port="1" />
-		<edge from-layer="1013" from-port="2" to-layer="1014" to-port="0" />
-		<edge from-layer="1014" from-port="1" to-layer="1015" to-port="1" />
-		<edge from-layer="1015" from-port="2" to-layer="1018" to-port="0" />
-		<edge from-layer="1016" from-port="0" to-layer="1017" to-port="0" />
-		<edge from-layer="1017" from-port="1" to-layer="1018" to-port="1" />
-		<edge from-layer="1018" from-port="2" to-layer="1019" to-port="1" />
-		<edge from-layer="1019" from-port="2" to-layer="1020" to-port="0" />
-		<edge from-layer="1020" from-port="2" to-layer="1022" to-port="0" />
-		<edge from-layer="1020" from-port="2" to-layer="1262" to-port="1" />
-		<edge from-layer="1021" from-port="0" to-layer="1022" to-port="1" />
-		<edge from-layer="1022" from-port="2" to-layer="1025" to-port="0" />
-		<edge from-layer="1022" from-port="2" to-layer="1056" to-port="0" />
-		<edge from-layer="1023" from-port="0" to-layer="1024" to-port="0" />
-		<edge from-layer="1024" from-port="1" to-layer="1025" to-port="1" />
-		<edge from-layer="1025" from-port="2" to-layer="1028" to-port="0" />
-		<edge from-layer="1026" from-port="0" to-layer="1027" to-port="0" />
-		<edge from-layer="1027" from-port="1" to-layer="1028" to-port="1" />
-		<edge from-layer="1028" from-port="2" to-layer="1037" to-port="0" />
-		<edge from-layer="1029" from-port="0" to-layer="1030" to-port="0" />
-		<edge from-layer="1030" from-port="1" to-layer="1033" to-port="0" />
-		<edge from-layer="1031" from-port="0" to-layer="1032" to-port="0" />
-		<edge from-layer="1032" from-port="1" to-layer="1033" to-port="1" />
-		<edge from-layer="1033" from-port="2" to-layer="1035" to-port="0" />
-		<edge from-layer="1034" from-port="0" to-layer="1035" to-port="1" />
-		<edge from-layer="1035" from-port="2" to-layer="1036" to-port="0" />
-		<edge from-layer="1036" from-port="1" to-layer="1037" to-port="1" />
-		<edge from-layer="1037" from-port="2" to-layer="1040" to-port="0" />
-		<edge from-layer="1038" from-port="0" to-layer="1039" to-port="0" />
-		<edge from-layer="1039" from-port="1" to-layer="1040" to-port="1" />
-		<edge from-layer="1040" from-port="2" to-layer="1041" to-port="0" />
-		<edge from-layer="1041" from-port="1" to-layer="1050" to-port="0" />
-		<edge from-layer="1042" from-port="0" to-layer="1043" to-port="0" />
-		<edge from-layer="1043" from-port="1" to-layer="1046" to-port="0" />
-		<edge from-layer="1044" from-port="0" to-layer="1045" to-port="0" />
-		<edge from-layer="1045" from-port="1" to-layer="1046" to-port="1" />
-		<edge from-layer="1046" from-port="2" to-layer="1048" to-port="0" />
-		<edge from-layer="1047" from-port="0" to-layer="1048" to-port="1" />
-		<edge from-layer="1048" from-port="2" to-layer="1049" to-port="0" />
-		<edge from-layer="1049" from-port="1" to-layer="1050" to-port="1" />
-		<edge from-layer="1050" from-port="2" to-layer="1053" to-port="0" />
-		<edge from-layer="1051" from-port="0" to-layer="1052" to-port="0" />
-		<edge from-layer="1052" from-port="1" to-layer="1053" to-port="1" />
-		<edge from-layer="1053" from-port="2" to-layer="1261" to-port="0" />
-		<edge from-layer="1054" from-port="0" to-layer="1055" to-port="0" />
-		<edge from-layer="1055" from-port="1" to-layer="1056" to-port="1" />
-		<edge from-layer="1056" from-port="2" to-layer="1059" to-port="0" />
-		<edge from-layer="1057" from-port="0" to-layer="1058" to-port="0" />
-		<edge from-layer="1058" from-port="1" to-layer="1059" to-port="1" />
-		<edge from-layer="1059" from-port="2" to-layer="1068" to-port="0" />
-		<edge from-layer="1060" from-port="0" to-layer="1061" to-port="0" />
-		<edge from-layer="1061" from-port="1" to-layer="1064" to-port="0" />
-		<edge from-layer="1062" from-port="0" to-layer="1063" to-port="0" />
-		<edge from-layer="1063" from-port="1" to-layer="1064" to-port="1" />
-		<edge from-layer="1064" from-port="2" to-layer="1066" to-port="0" />
-		<edge from-layer="1065" from-port="0" to-layer="1066" to-port="1" />
-		<edge from-layer="1066" from-port="2" to-layer="1067" to-port="0" />
-		<edge from-layer="1067" from-port="1" to-layer="1068" to-port="1" />
-		<edge from-layer="1068" from-port="2" to-layer="1071" to-port="0" />
-		<edge from-layer="1069" from-port="0" to-layer="1070" to-port="0" />
-		<edge from-layer="1070" from-port="1" to-layer="1071" to-port="1" />
-		<edge from-layer="1071" from-port="2" to-layer="1073" to-port="0" />
-		<edge from-layer="1072" from-port="0" to-layer="1073" to-port="1" />
-		<edge from-layer="1073" from-port="2" to-layer="1090" to-port="0" />
-		<edge from-layer="1073" from-port="2" to-layer="1077" to-port="0" />
-		<edge from-layer="1073" from-port="2" to-layer="1201" to-port="0" />
-		<edge from-layer="1074" from-port="0" to-layer="1077" to-port="1" />
-		<edge from-layer="1075" from-port="0" to-layer="1077" to-port="2" />
-		<edge from-layer="1076" from-port="0" to-layer="1077" to-port="3" />
-		<edge from-layer="1077" from-port="4" to-layer="1079" to-port="0" />
-		<edge from-layer="1078" from-port="0" to-layer="1079" to-port="1" />
-		<edge from-layer="1079" from-port="2" to-layer="1152" to-port="0" />
-		<edge from-layer="1079" from-port="2" to-layer="1083" to-port="0" />
-		<edge from-layer="1080" from-port="0" to-layer="1083" to-port="1" />
-		<edge from-layer="1081" from-port="0" to-layer="1083" to-port="2" />
-		<edge from-layer="1082" from-port="0" to-layer="1083" to-port="3" />
-		<edge from-layer="1083" from-port="4" to-layer="1117" to-port="0" />
-		<edge from-layer="1083" from-port="4" to-layer="1120" to-port="0" />
-		<edge from-layer="1083" from-port="4" to-layer="1131" to-port="0" />
-		<edge from-layer="1083" from-port="4" to-layer="1137" to-port="0" />
-		<edge from-layer="1084" from-port="0" to-layer="1085" to-port="0" />
-		<edge from-layer="1085" from-port="1" to-layer="1113" to-port="0" />
-		<edge from-layer="1086" from-port="0" to-layer="1113" to-port="1" />
-		<edge from-layer="1087" from-port="0" to-layer="1090" to-port="1" />
-		<edge from-layer="1088" from-port="0" to-layer="1090" to-port="2" />
-		<edge from-layer="1089" from-port="0" to-layer="1090" to-port="3" />
-		<edge from-layer="1090" from-port="4" to-layer="1092" to-port="0" />
-		<edge from-layer="1091" from-port="0" to-layer="1092" to-port="1" />
-		<edge from-layer="1092" from-port="2" to-layer="1185" to-port="0" />
-		<edge from-layer="1092" from-port="2" to-layer="1157" to-port="0" />
-		<edge from-layer="1092" from-port="2" to-layer="1093" to-port="0" />
-		<edge from-layer="1093" from-port="1" to-layer="1095" to-port="0" />
-		<edge from-layer="1094" from-port="0" to-layer="1095" to-port="2" />
-		<edge from-layer="1095" from-port="3" to-layer="1109" to-port="0" />
+		<edge from-layer="933" from-port="2" to-layer="935" to-port="0" />
+		<edge from-layer="934" from-port="0" to-layer="935" to-port="1" />
+		<edge from-layer="935" from-port="2" to-layer="936" to-port="0" />
+		<edge from-layer="936" from-port="1" to-layer="937" to-port="1" />
+		<edge from-layer="937" from-port="2" to-layer="940" to-port="0" />
+		<edge from-layer="938" from-port="0" to-layer="939" to-port="0" />
+		<edge from-layer="939" from-port="1" to-layer="940" to-port="1" />
+		<edge from-layer="940" from-port="2" to-layer="941" to-port="0" />
+		<edge from-layer="941" from-port="1" to-layer="950" to-port="0" />
+		<edge from-layer="942" from-port="0" to-layer="943" to-port="0" />
+		<edge from-layer="943" from-port="1" to-layer="946" to-port="0" />
+		<edge from-layer="944" from-port="0" to-layer="945" to-port="0" />
+		<edge from-layer="945" from-port="1" to-layer="946" to-port="1" />
+		<edge from-layer="946" from-port="2" to-layer="948" to-port="0" />
+		<edge from-layer="947" from-port="0" to-layer="948" to-port="1" />
+		<edge from-layer="948" from-port="2" to-layer="949" to-port="0" />
+		<edge from-layer="949" from-port="1" to-layer="950" to-port="1" />
+		<edge from-layer="950" from-port="2" to-layer="953" to-port="0" />
+		<edge from-layer="951" from-port="0" to-layer="952" to-port="0" />
+		<edge from-layer="952" from-port="1" to-layer="953" to-port="1" />
+		<edge from-layer="953" from-port="2" to-layer="1079" to-port="0" />
+		<edge from-layer="954" from-port="0" to-layer="955" to-port="1" />
+		<edge from-layer="955" from-port="2" to-layer="957" to-port="0" />
+		<edge from-layer="956" from-port="0" to-layer="957" to-port="1" />
+		<edge from-layer="957" from-port="2" to-layer="966" to-port="0" />
+		<edge from-layer="958" from-port="0" to-layer="959" to-port="0" />
+		<edge from-layer="959" from-port="1" to-layer="962" to-port="0" />
+		<edge from-layer="960" from-port="0" to-layer="961" to-port="0" />
+		<edge from-layer="961" from-port="1" to-layer="962" to-port="1" />
+		<edge from-layer="962" from-port="2" to-layer="964" to-port="0" />
+		<edge from-layer="963" from-port="0" to-layer="964" to-port="1" />
+		<edge from-layer="964" from-port="2" to-layer="965" to-port="0" />
+		<edge from-layer="965" from-port="1" to-layer="966" to-port="1" />
+		<edge from-layer="966" from-port="2" to-layer="968" to-port="0" />
+		<edge from-layer="967" from-port="0" to-layer="968" to-port="1" />
+		<edge from-layer="968" from-port="2" to-layer="970" to-port="0" />
+		<edge from-layer="969" from-port="0" to-layer="970" to-port="1" />
+		<edge from-layer="970" from-port="2" to-layer="1012" to-port="0" />
+		<edge from-layer="970" from-port="2" to-layer="974" to-port="0" />
+		<edge from-layer="970" from-port="2" to-layer="1051" to-port="0" />
+		<edge from-layer="971" from-port="0" to-layer="974" to-port="1" />
+		<edge from-layer="972" from-port="0" to-layer="974" to-port="2" />
+		<edge from-layer="973" from-port="0" to-layer="974" to-port="3" />
+		<edge from-layer="974" from-port="5" to-layer="976" to-port="0" />
+		<edge from-layer="975" from-port="0" to-layer="976" to-port="1" />
+		<edge from-layer="976" from-port="2" to-layer="998" to-port="0" />
+		<edge from-layer="976" from-port="2" to-layer="980" to-port="0" />
+		<edge from-layer="977" from-port="0" to-layer="980" to-port="1" />
+		<edge from-layer="978" from-port="0" to-layer="980" to-port="2" />
+		<edge from-layer="979" from-port="0" to-layer="980" to-port="3" />
+		<edge from-layer="980" from-port="5" to-layer="985" to-port="0" />
+		<edge from-layer="980" from-port="5" to-layer="991" to-port="0" />
+		<edge from-layer="980" from-port="5" to-layer="981" to-port="0" />
+		<edge from-layer="981" from-port="2" to-layer="994" to-port="0" />
+		<edge from-layer="982" from-port="0" to-layer="985" to-port="1" />
+		<edge from-layer="983" from-port="0" to-layer="985" to-port="2" />
+		<edge from-layer="984" from-port="0" to-layer="985" to-port="3" />
+		<edge from-layer="985" from-port="5" to-layer="987" to-port="0" />
+		<edge from-layer="986" from-port="0" to-layer="987" to-port="1" />
+		<edge from-layer="987" from-port="2" to-layer="992" to-port="0" />
+		<edge from-layer="988" from-port="0" to-layer="991" to-port="1" />
+		<edge from-layer="989" from-port="0" to-layer="991" to-port="2" />
+		<edge from-layer="990" from-port="0" to-layer="991" to-port="3" />
+		<edge from-layer="991" from-port="5" to-layer="992" to-port="1" />
+		<edge from-layer="992" from-port="2" to-layer="993" to-port="0" />
+		<edge from-layer="993" from-port="2" to-layer="994" to-port="1" />
+		<edge from-layer="994" from-port="2" to-layer="999" to-port="0" />
+		<edge from-layer="995" from-port="0" to-layer="998" to-port="1" />
+		<edge from-layer="996" from-port="0" to-layer="998" to-port="2" />
+		<edge from-layer="997" from-port="0" to-layer="998" to-port="3" />
+		<edge from-layer="998" from-port="5" to-layer="999" to-port="1" />
+		<edge from-layer="999" from-port="2" to-layer="1063" to-port="0" />
+		<edge from-layer="1000" from-port="0" to-layer="1005" to-port="0" />
+		<edge from-layer="1001" from-port="0" to-layer="1004" to-port="1" />
+		<edge from-layer="1002" from-port="0" to-layer="1004" to-port="2" />
+		<edge from-layer="1003" from-port="0" to-layer="1004" to-port="3" />
+		<edge from-layer="1004" from-port="4" to-layer="1005" to-port="1" />
+		<edge from-layer="1005" from-port="2" to-layer="1006" to-port="0" />
+		<edge from-layer="1006" from-port="1" to-layer="1008" to-port="0" />
+		<edge from-layer="1007" from-port="0" to-layer="1008" to-port="2" />
+		<edge from-layer="1008" from-port="3" to-layer="1038" to-port="0" />
+		<edge from-layer="1008" from-port="3" to-layer="1056" to-port="0" />
+		<edge from-layer="1009" from-port="0" to-layer="1012" to-port="1" />
+		<edge from-layer="1010" from-port="0" to-layer="1012" to-port="2" />
+		<edge from-layer="1011" from-port="0" to-layer="1012" to-port="3" />
+		<edge from-layer="1012" from-port="5" to-layer="1014" to-port="0" />
+		<edge from-layer="1013" from-port="0" to-layer="1014" to-port="1" />
+		<edge from-layer="1014" from-port="2" to-layer="1018" to-port="0" />
+		<edge from-layer="1014" from-port="2" to-layer="1036" to-port="0" />
+		<edge from-layer="1015" from-port="0" to-layer="1018" to-port="1" />
+		<edge from-layer="1016" from-port="0" to-layer="1018" to-port="2" />
+		<edge from-layer="1017" from-port="0" to-layer="1018" to-port="3" />
+		<edge from-layer="1018" from-port="5" to-layer="1023" to-port="0" />
+		<edge from-layer="1018" from-port="5" to-layer="1019" to-port="0" />
+		<edge from-layer="1018" from-port="5" to-layer="1029" to-port="0" />
+		<edge from-layer="1019" from-port="2" to-layer="1032" to-port="0" />
+		<edge from-layer="1020" from-port="0" to-layer="1023" to-port="1" />
+		<edge from-layer="1021" from-port="0" to-layer="1023" to-port="2" />
+		<edge from-layer="1022" from-port="0" to-layer="1023" to-port="3" />
+		<edge from-layer="1023" from-port="5" to-layer="1025" to-port="0" />
+		<edge from-layer="1024" from-port="0" to-layer="1025" to-port="1" />
+		<edge from-layer="1025" from-port="2" to-layer="1030" to-port="0" />
+		<edge from-layer="1026" from-port="0" to-layer="1029" to-port="1" />
+		<edge from-layer="1027" from-port="0" to-layer="1029" to-port="2" />
+		<edge from-layer="1028" from-port="0" to-layer="1029" to-port="3" />
+		<edge from-layer="1029" from-port="5" to-layer="1030" to-port="1" />
+		<edge from-layer="1030" from-port="2" to-layer="1031" to-port="0" />
+		<edge from-layer="1031" from-port="2" to-layer="1032" to-port="1" />
+		<edge from-layer="1032" from-port="2" to-layer="1037" to-port="0" />
+		<edge from-layer="1033" from-port="0" to-layer="1036" to-port="1" />
+		<edge from-layer="1034" from-port="0" to-layer="1036" to-port="2" />
+		<edge from-layer="1035" from-port="0" to-layer="1036" to-port="3" />
+		<edge from-layer="1036" from-port="5" to-layer="1037" to-port="1" />
+		<edge from-layer="1037" from-port="2" to-layer="1038" to-port="1" />
+		<edge from-layer="1038" from-port="2" to-layer="1629" to-port="0" />
+		<edge from-layer="1038" from-port="2" to-layer="1063" to-port="1" />
+		<edge from-layer="1039" from-port="0" to-layer="1044" to-port="0" />
+		<edge from-layer="1040" from-port="0" to-layer="1043" to-port="1" />
+		<edge from-layer="1041" from-port="0" to-layer="1043" to-port="2" />
+		<edge from-layer="1042" from-port="0" to-layer="1043" to-port="3" />
+		<edge from-layer="1043" from-port="4" to-layer="1044" to-port="1" />
+		<edge from-layer="1044" from-port="2" to-layer="1045" to-port="0" />
+		<edge from-layer="1045" from-port="1" to-layer="1047" to-port="0" />
+		<edge from-layer="1046" from-port="0" to-layer="1047" to-port="2" />
+		<edge from-layer="1047" from-port="3" to-layer="1054" to-port="0" />
+		<edge from-layer="1048" from-port="0" to-layer="1051" to-port="1" />
+		<edge from-layer="1049" from-port="0" to-layer="1051" to-port="2" />
+		<edge from-layer="1050" from-port="0" to-layer="1051" to-port="3" />
+		<edge from-layer="1051" from-port="5" to-layer="1053" to-port="0" />
+		<edge from-layer="1052" from-port="0" to-layer="1053" to-port="1" />
+		<edge from-layer="1053" from-port="2" to-layer="1054" to-port="1" />
+		<edge from-layer="1054" from-port="2" to-layer="1628" to-port="0" />
+		<edge from-layer="1054" from-port="2" to-layer="1063" to-port="2" />
+		<edge from-layer="1055" from-port="0" to-layer="1062" to-port="1" />
+		<edge from-layer="1056" from-port="1" to-layer="1059" to-port="0" />
+		<edge from-layer="1057" from-port="0" to-layer="1059" to-port="1" />
+		<edge from-layer="1058" from-port="0" to-layer="1059" to-port="2" />
+		<edge from-layer="1059" from-port="3" to-layer="1060" to-port="0" />
+		<edge from-layer="1060" from-port="2" to-layer="1062" to-port="2" />
+		<edge from-layer="1061" from-port="0" to-layer="1062" to-port="3" />
+		<edge from-layer="1062" from-port="5" to-layer="1063" to-port="3" />
+		<edge from-layer="1063" from-port="4" to-layer="1065" to-port="0" />
+		<edge from-layer="1064" from-port="0" to-layer="1065" to-port="1" />
+		<edge from-layer="1065" from-port="2" to-layer="1067" to-port="0" />
+		<edge from-layer="1066" from-port="0" to-layer="1067" to-port="1" />
+		<edge from-layer="1067" from-port="2" to-layer="1076" to-port="0" />
+		<edge from-layer="1068" from-port="0" to-layer="1069" to-port="0" />
+		<edge from-layer="1069" from-port="1" to-layer="1072" to-port="0" />
+		<edge from-layer="1070" from-port="0" to-layer="1071" to-port="0" />
+		<edge from-layer="1071" from-port="1" to-layer="1072" to-port="1" />
+		<edge from-layer="1072" from-port="2" to-layer="1074" to-port="0" />
+		<edge from-layer="1073" from-port="0" to-layer="1074" to-port="1" />
+		<edge from-layer="1074" from-port="2" to-layer="1075" to-port="0" />
+		<edge from-layer="1075" from-port="1" to-layer="1076" to-port="1" />
+		<edge from-layer="1076" from-port="2" to-layer="1078" to-port="0" />
+		<edge from-layer="1077" from-port="0" to-layer="1078" to-port="1" />
+		<edge from-layer="1078" from-port="2" to-layer="1079" to-port="1" />
+		<edge from-layer="1079" from-port="2" to-layer="1080" to-port="0" />
+		<edge from-layer="1080" from-port="2" to-layer="1082" to-port="0" />
+		<edge from-layer="1080" from-port="2" to-layer="1238" to-port="1" />
+		<edge from-layer="1081" from-port="0" to-layer="1082" to-port="1" />
+		<edge from-layer="1082" from-port="2" to-layer="1113" to-port="0" />
+		<edge from-layer="1082" from-port="2" to-layer="1084" to-port="0" />
+		<edge from-layer="1083" from-port="0" to-layer="1084" to-port="1" />
+		<edge from-layer="1084" from-port="2" to-layer="1086" to-port="0" />
+		<edge from-layer="1085" from-port="0" to-layer="1086" to-port="1" />
+		<edge from-layer="1086" from-port="2" to-layer="1095" to-port="0" />
+		<edge from-layer="1087" from-port="0" to-layer="1088" to-port="0" />
+		<edge from-layer="1088" from-port="1" to-layer="1091" to-port="0" />
+		<edge from-layer="1089" from-port="0" to-layer="1090" to-port="0" />
+		<edge from-layer="1090" from-port="1" to-layer="1091" to-port="1" />
+		<edge from-layer="1091" from-port="2" to-layer="1093" to-port="0" />
+		<edge from-layer="1092" from-port="0" to-layer="1093" to-port="1" />
+		<edge from-layer="1093" from-port="2" to-layer="1094" to-port="0" />
+		<edge from-layer="1094" from-port="1" to-layer="1095" to-port="1" />
+		<edge from-layer="1095" from-port="2" to-layer="1098" to-port="0" />
 		<edge from-layer="1096" from-port="0" to-layer="1097" to-port="0" />
-		<edge from-layer="1097" from-port="1" to-layer="1102" to-port="0" />
-		<edge from-layer="1098" from-port="0" to-layer="1101" to-port="1" />
-		<edge from-layer="1099" from-port="0" to-layer="1101" to-port="2" />
-		<edge from-layer="1100" from-port="0" to-layer="1101" to-port="3" />
-		<edge from-layer="1101" from-port="4" to-layer="1102" to-port="1" />
-		<edge from-layer="1102" from-port="2" to-layer="1103" to-port="0" />
-		<edge from-layer="1103" from-port="1" to-layer="1105" to-port="0" />
-		<edge from-layer="1104" from-port="0" to-layer="1105" to-port="2" />
-		<edge from-layer="1105" from-port="3" to-layer="1106" to-port="0" />
-		<edge from-layer="1105" from-port="3" to-layer="1187" to-port="0" />
-		<edge from-layer="1106" from-port="1" to-layer="1108" to-port="0" />
-		<edge from-layer="1107" from-port="0" to-layer="1108" to-port="2" />
-		<edge from-layer="1108" from-port="3" to-layer="1109" to-port="1" />
-		<edge from-layer="1109" from-port="2" to-layer="1111" to-port="0" />
-		<edge from-layer="1110" from-port="0" to-layer="1111" to-port="1" />
-		<edge from-layer="1111" from-port="2" to-layer="1143" to-port="2" />
-		<edge from-layer="1111" from-port="2" to-layer="1113" to-port="2" />
-		<edge from-layer="1112" from-port="0" to-layer="1113" to-port="3" />
-		<edge from-layer="1113" from-port="4" to-layer="1115" to-port="0" />
-		<edge from-layer="1114" from-port="0" to-layer="1115" to-port="2" />
-		<edge from-layer="1115" from-port="3" to-layer="1116" to-port="0" />
-		<edge from-layer="1116" from-port="2" to-layer="1158" to-port="1" />
-		<edge from-layer="1116" from-port="2" to-layer="1117" to-port="1" />
-		<edge from-layer="1117" from-port="2" to-layer="1148" to-port="0" />
-		<edge from-layer="1118" from-port="0" to-layer="1128" to-port="0" />
-		<edge from-layer="1119" from-port="0" to-layer="1128" to-port="1" />
-		<edge from-layer="1120" from-port="1" to-layer="1122" to-port="0" />
-		<edge from-layer="1121" from-port="0" to-layer="1122" to-port="2" />
-		<edge from-layer="1122" from-port="3" to-layer="1123" to-port="0" />
-		<edge from-layer="1123" from-port="2" to-layer="1124" to-port="0" />
-		<edge from-layer="1124" from-port="1" to-layer="1126" to-port="0" />
+		<edge from-layer="1097" from-port="1" to-layer="1098" to-port="1" />
+		<edge from-layer="1098" from-port="2" to-layer="1099" to-port="0" />
+		<edge from-layer="1099" from-port="1" to-layer="1108" to-port="0" />
+		<edge from-layer="1100" from-port="0" to-layer="1101" to-port="0" />
+		<edge from-layer="1101" from-port="1" to-layer="1104" to-port="0" />
+		<edge from-layer="1102" from-port="0" to-layer="1103" to-port="0" />
+		<edge from-layer="1103" from-port="1" to-layer="1104" to-port="1" />
+		<edge from-layer="1104" from-port="2" to-layer="1106" to-port="0" />
+		<edge from-layer="1105" from-port="0" to-layer="1106" to-port="1" />
+		<edge from-layer="1106" from-port="2" to-layer="1107" to-port="0" />
+		<edge from-layer="1107" from-port="1" to-layer="1108" to-port="1" />
+		<edge from-layer="1108" from-port="2" to-layer="1111" to-port="0" />
+		<edge from-layer="1109" from-port="0" to-layer="1110" to-port="0" />
+		<edge from-layer="1110" from-port="1" to-layer="1111" to-port="1" />
+		<edge from-layer="1111" from-port="2" to-layer="1237" to-port="0" />
+		<edge from-layer="1112" from-port="0" to-layer="1113" to-port="1" />
+		<edge from-layer="1113" from-port="2" to-layer="1115" to-port="0" />
+		<edge from-layer="1114" from-port="0" to-layer="1115" to-port="1" />
+		<edge from-layer="1115" from-port="2" to-layer="1124" to-port="0" />
+		<edge from-layer="1116" from-port="0" to-layer="1117" to-port="0" />
+		<edge from-layer="1117" from-port="1" to-layer="1120" to-port="0" />
+		<edge from-layer="1118" from-port="0" to-layer="1119" to-port="0" />
+		<edge from-layer="1119" from-port="1" to-layer="1120" to-port="1" />
+		<edge from-layer="1120" from-port="2" to-layer="1122" to-port="0" />
+		<edge from-layer="1121" from-port="0" to-layer="1122" to-port="1" />
+		<edge from-layer="1122" from-port="2" to-layer="1123" to-port="0" />
+		<edge from-layer="1123" from-port="1" to-layer="1124" to-port="1" />
+		<edge from-layer="1124" from-port="2" to-layer="1126" to-port="0" />
 		<edge from-layer="1125" from-port="0" to-layer="1126" to-port="1" />
-		<edge from-layer="1126" from-port="2" to-layer="1128" to-port="2" />
-		<edge from-layer="1127" from-port="0" to-layer="1128" to-port="3" />
-		<edge from-layer="1128" from-port="4" to-layer="1137" to-port="2" />
-		<edge from-layer="1128" from-port="4" to-layer="1131" to-port="1" />
-		<edge from-layer="1129" from-port="0" to-layer="1131" to-port="2" />
-		<edge from-layer="1130" from-port="0" to-layer="1131" to-port="3" />
-		<edge from-layer="1131" from-port="4" to-layer="1134" to-port="0" />
-		<edge from-layer="1132" from-port="0" to-layer="1133" to-port="0" />
-		<edge from-layer="1133" from-port="1" to-layer="1134" to-port="1" />
+		<edge from-layer="1126" from-port="2" to-layer="1128" to-port="0" />
+		<edge from-layer="1127" from-port="0" to-layer="1128" to-port="1" />
+		<edge from-layer="1128" from-port="2" to-layer="1170" to-port="0" />
+		<edge from-layer="1128" from-port="2" to-layer="1209" to-port="0" />
+		<edge from-layer="1128" from-port="2" to-layer="1132" to-port="0" />
+		<edge from-layer="1129" from-port="0" to-layer="1132" to-port="1" />
+		<edge from-layer="1130" from-port="0" to-layer="1132" to-port="2" />
+		<edge from-layer="1131" from-port="0" to-layer="1132" to-port="3" />
+		<edge from-layer="1132" from-port="5" to-layer="1134" to-port="0" />
+		<edge from-layer="1133" from-port="0" to-layer="1134" to-port="1" />
 		<edge from-layer="1134" from-port="2" to-layer="1138" to-port="0" />
-		<edge from-layer="1135" from-port="0" to-layer="1137" to-port="1" />
-		<edge from-layer="1136" from-port="0" to-layer="1137" to-port="3" />
-		<edge from-layer="1137" from-port="4" to-layer="1138" to-port="1" />
-		<edge from-layer="1138" from-port="2" to-layer="1147" to-port="0" />
-		<edge from-layer="1139" from-port="0" to-layer="1140" to-port="0" />
-		<edge from-layer="1140" from-port="1" to-layer="1143" to-port="0" />
-		<edge from-layer="1141" from-port="0" to-layer="1143" to-port="1" />
+		<edge from-layer="1134" from-port="2" to-layer="1156" to-port="0" />
+		<edge from-layer="1135" from-port="0" to-layer="1138" to-port="1" />
+		<edge from-layer="1136" from-port="0" to-layer="1138" to-port="2" />
+		<edge from-layer="1137" from-port="0" to-layer="1138" to-port="3" />
+		<edge from-layer="1138" from-port="5" to-layer="1139" to-port="0" />
+		<edge from-layer="1138" from-port="5" to-layer="1143" to-port="0" />
+		<edge from-layer="1138" from-port="5" to-layer="1149" to-port="0" />
+		<edge from-layer="1139" from-port="2" to-layer="1152" to-port="0" />
+		<edge from-layer="1140" from-port="0" to-layer="1143" to-port="1" />
+		<edge from-layer="1141" from-port="0" to-layer="1143" to-port="2" />
 		<edge from-layer="1142" from-port="0" to-layer="1143" to-port="3" />
-		<edge from-layer="1143" from-port="4" to-layer="1145" to-port="0" />
-		<edge from-layer="1144" from-port="0" to-layer="1145" to-port="2" />
-		<edge from-layer="1145" from-port="3" to-layer="1146" to-port="0" />
-		<edge from-layer="1146" from-port="2" to-layer="1180" to-port="1" />
-		<edge from-layer="1146" from-port="2" to-layer="1147" to-port="1" />
-		<edge from-layer="1147" from-port="2" to-layer="1148" to-port="1" />
-		<edge from-layer="1148" from-port="2" to-layer="1153" to-port="0" />
-		<edge from-layer="1149" from-port="0" to-layer="1152" to-port="1" />
-		<edge from-layer="1150" from-port="0" to-layer="1152" to-port="2" />
-		<edge from-layer="1151" from-port="0" to-layer="1152" to-port="3" />
-		<edge from-layer="1152" from-port="4" to-layer="1153" to-port="1" />
-		<edge from-layer="1153" from-port="2" to-layer="1244" to-port="0" />
-		<edge from-layer="1153" from-port="2" to-layer="1211" to-port="0" />
-		<edge from-layer="1154" from-port="0" to-layer="1157" to-port="1" />
-		<edge from-layer="1155" from-port="0" to-layer="1157" to-port="2" />
-		<edge from-layer="1156" from-port="0" to-layer="1157" to-port="3" />
-		<edge from-layer="1157" from-port="4" to-layer="1161" to-port="0" />
-		<edge from-layer="1157" from-port="4" to-layer="1158" to-port="0" />
-		<edge from-layer="1157" from-port="4" to-layer="1172" to-port="0" />
-		<edge from-layer="1157" from-port="4" to-layer="1178" to-port="0" />
-		<edge from-layer="1158" from-port="2" to-layer="1181" to-port="0" />
-		<edge from-layer="1159" from-port="0" to-layer="1169" to-port="0" />
-		<edge from-layer="1160" from-port="0" to-layer="1169" to-port="1" />
-		<edge from-layer="1161" from-port="1" to-layer="1163" to-port="0" />
-		<edge from-layer="1162" from-port="0" to-layer="1163" to-port="2" />
-		<edge from-layer="1163" from-port="3" to-layer="1164" to-port="0" />
-		<edge from-layer="1164" from-port="2" to-layer="1165" to-port="0" />
-		<edge from-layer="1165" from-port="1" to-layer="1167" to-port="0" />
-		<edge from-layer="1166" from-port="0" to-layer="1167" to-port="1" />
-		<edge from-layer="1167" from-port="2" to-layer="1169" to-port="2" />
-		<edge from-layer="1168" from-port="0" to-layer="1169" to-port="3" />
-		<edge from-layer="1169" from-port="4" to-layer="1178" to-port="2" />
-		<edge from-layer="1169" from-port="4" to-layer="1172" to-port="1" />
-		<edge from-layer="1170" from-port="0" to-layer="1172" to-port="2" />
-		<edge from-layer="1171" from-port="0" to-layer="1172" to-port="3" />
-		<edge from-layer="1172" from-port="4" to-layer="1175" to-port="0" />
-		<edge from-layer="1173" from-port="0" to-layer="1174" to-port="0" />
-		<edge from-layer="1174" from-port="1" to-layer="1175" to-port="1" />
-		<edge from-layer="1175" from-port="2" to-layer="1179" to-port="0" />
-		<edge from-layer="1176" from-port="0" to-layer="1178" to-port="1" />
-		<edge from-layer="1177" from-port="0" to-layer="1178" to-port="3" />
-		<edge from-layer="1178" from-port="4" to-layer="1179" to-port="1" />
-		<edge from-layer="1179" from-port="2" to-layer="1180" to-port="0" />
-		<edge from-layer="1180" from-port="2" to-layer="1181" to-port="1" />
-		<edge from-layer="1181" from-port="2" to-layer="1186" to-port="0" />
-		<edge from-layer="1182" from-port="0" to-layer="1185" to-port="1" />
-		<edge from-layer="1183" from-port="0" to-layer="1185" to-port="2" />
-		<edge from-layer="1184" from-port="0" to-layer="1185" to-port="3" />
-		<edge from-layer="1185" from-port="4" to-layer="1186" to-port="1" />
-		<edge from-layer="1186" from-port="2" to-layer="1187" to-port="1" />
-		<edge from-layer="1187" from-port="2" to-layer="1208" to-port="0" />
-		<edge from-layer="1187" from-port="2" to-layer="1244" to-port="1" />
-		<edge from-layer="1187" from-port="2" to-layer="2346" to-port="0" />
-		<edge from-layer="1188" from-port="0" to-layer="1189" to-port="0" />
-		<edge from-layer="1189" from-port="1" to-layer="1194" to-port="0" />
-		<edge from-layer="1190" from-port="0" to-layer="1193" to-port="1" />
-		<edge from-layer="1191" from-port="0" to-layer="1193" to-port="2" />
-		<edge from-layer="1192" from-port="0" to-layer="1193" to-port="3" />
-		<edge from-layer="1193" from-port="4" to-layer="1194" to-port="1" />
-		<edge from-layer="1194" from-port="2" to-layer="1195" to-port="0" />
-		<edge from-layer="1195" from-port="1" to-layer="1197" to-port="0" />
-		<edge from-layer="1196" from-port="0" to-layer="1197" to-port="2" />
-		<edge from-layer="1197" from-port="3" to-layer="1204" to-port="0" />
+		<edge from-layer="1143" from-port="5" to-layer="1145" to-port="0" />
+		<edge from-layer="1144" from-port="0" to-layer="1145" to-port="1" />
+		<edge from-layer="1145" from-port="2" to-layer="1150" to-port="0" />
+		<edge from-layer="1146" from-port="0" to-layer="1149" to-port="1" />
+		<edge from-layer="1147" from-port="0" to-layer="1149" to-port="2" />
+		<edge from-layer="1148" from-port="0" to-layer="1149" to-port="3" />
+		<edge from-layer="1149" from-port="5" to-layer="1150" to-port="1" />
+		<edge from-layer="1150" from-port="2" to-layer="1151" to-port="0" />
+		<edge from-layer="1151" from-port="2" to-layer="1152" to-port="1" />
+		<edge from-layer="1152" from-port="2" to-layer="1157" to-port="0" />
+		<edge from-layer="1153" from-port="0" to-layer="1156" to-port="1" />
+		<edge from-layer="1154" from-port="0" to-layer="1156" to-port="2" />
+		<edge from-layer="1155" from-port="0" to-layer="1156" to-port="3" />
+		<edge from-layer="1156" from-port="5" to-layer="1157" to-port="1" />
+		<edge from-layer="1157" from-port="2" to-layer="1221" to-port="0" />
+		<edge from-layer="1158" from-port="0" to-layer="1163" to-port="0" />
+		<edge from-layer="1159" from-port="0" to-layer="1162" to-port="1" />
+		<edge from-layer="1160" from-port="0" to-layer="1162" to-port="2" />
+		<edge from-layer="1161" from-port="0" to-layer="1162" to-port="3" />
+		<edge from-layer="1162" from-port="4" to-layer="1163" to-port="1" />
+		<edge from-layer="1163" from-port="2" to-layer="1164" to-port="0" />
+		<edge from-layer="1164" from-port="1" to-layer="1166" to-port="0" />
+		<edge from-layer="1165" from-port="0" to-layer="1166" to-port="2" />
+		<edge from-layer="1166" from-port="3" to-layer="1196" to-port="0" />
+		<edge from-layer="1166" from-port="3" to-layer="1214" to-port="0" />
+		<edge from-layer="1167" from-port="0" to-layer="1170" to-port="1" />
+		<edge from-layer="1168" from-port="0" to-layer="1170" to-port="2" />
+		<edge from-layer="1169" from-port="0" to-layer="1170" to-port="3" />
+		<edge from-layer="1170" from-port="5" to-layer="1172" to-port="0" />
+		<edge from-layer="1171" from-port="0" to-layer="1172" to-port="1" />
+		<edge from-layer="1172" from-port="2" to-layer="1176" to-port="0" />
+		<edge from-layer="1172" from-port="2" to-layer="1194" to-port="0" />
+		<edge from-layer="1173" from-port="0" to-layer="1176" to-port="1" />
+		<edge from-layer="1174" from-port="0" to-layer="1176" to-port="2" />
+		<edge from-layer="1175" from-port="0" to-layer="1176" to-port="3" />
+		<edge from-layer="1176" from-port="5" to-layer="1187" to-port="0" />
+		<edge from-layer="1176" from-port="5" to-layer="1181" to-port="0" />
+		<edge from-layer="1176" from-port="5" to-layer="1177" to-port="0" />
+		<edge from-layer="1177" from-port="2" to-layer="1190" to-port="0" />
+		<edge from-layer="1178" from-port="0" to-layer="1181" to-port="1" />
+		<edge from-layer="1179" from-port="0" to-layer="1181" to-port="2" />
+		<edge from-layer="1180" from-port="0" to-layer="1181" to-port="3" />
+		<edge from-layer="1181" from-port="5" to-layer="1183" to-port="0" />
+		<edge from-layer="1182" from-port="0" to-layer="1183" to-port="1" />
+		<edge from-layer="1183" from-port="2" to-layer="1188" to-port="0" />
+		<edge from-layer="1184" from-port="0" to-layer="1187" to-port="1" />
+		<edge from-layer="1185" from-port="0" to-layer="1187" to-port="2" />
+		<edge from-layer="1186" from-port="0" to-layer="1187" to-port="3" />
+		<edge from-layer="1187" from-port="5" to-layer="1188" to-port="1" />
+		<edge from-layer="1188" from-port="2" to-layer="1189" to-port="0" />
+		<edge from-layer="1189" from-port="2" to-layer="1190" to-port="1" />
+		<edge from-layer="1190" from-port="2" to-layer="1195" to-port="0" />
+		<edge from-layer="1191" from-port="0" to-layer="1194" to-port="1" />
+		<edge from-layer="1192" from-port="0" to-layer="1194" to-port="2" />
+		<edge from-layer="1193" from-port="0" to-layer="1194" to-port="3" />
+		<edge from-layer="1194" from-port="5" to-layer="1195" to-port="1" />
+		<edge from-layer="1195" from-port="2" to-layer="1196" to-port="1" />
+		<edge from-layer="1196" from-port="2" to-layer="1627" to-port="0" />
+		<edge from-layer="1196" from-port="2" to-layer="1221" to-port="1" />
+		<edge from-layer="1197" from-port="0" to-layer="1202" to-port="0" />
 		<edge from-layer="1198" from-port="0" to-layer="1201" to-port="1" />
 		<edge from-layer="1199" from-port="0" to-layer="1201" to-port="2" />
 		<edge from-layer="1200" from-port="0" to-layer="1201" to-port="3" />
-		<edge from-layer="1201" from-port="4" to-layer="1203" to-port="0" />
-		<edge from-layer="1202" from-port="0" to-layer="1203" to-port="1" />
-		<edge from-layer="1203" from-port="2" to-layer="1204" to-port="1" />
-		<edge from-layer="1204" from-port="2" to-layer="2345" to-port="0" />
-		<edge from-layer="1204" from-port="2" to-layer="1244" to-port="2" />
-		<edge from-layer="1205" from-port="0" to-layer="1223" to-port="0" />
-		<edge from-layer="1206" from-port="0" to-layer="1218" to-port="0" />
-		<edge from-layer="1207" from-port="0" to-layer="1218" to-port="1" />
-		<edge from-layer="1207" from-port="0" to-layer="1221" to-port="1" />
-		<edge from-layer="1208" from-port="1" to-layer="1210" to-port="0" />
-		<edge from-layer="1209" from-port="0" to-layer="1210" to-port="2" />
-		<edge from-layer="1210" from-port="3" to-layer="1220" to-port="0" />
-		<edge from-layer="1210" from-port="3" to-layer="1214" to-port="0" />
-		<edge from-layer="1211" from-port="1" to-layer="1213" to-port="0" />
-		<edge from-layer="1211" from-port="1" to-layer="1239" to-port="0" />
-		<edge from-layer="1212" from-port="0" to-layer="1213" to-port="2" />
-		<edge from-layer="1213" from-port="3" to-layer="1214" to-port="1" />
-		<edge from-layer="1214" from-port="2" to-layer="1216" to-port="0" />
-		<edge from-layer="1215" from-port="0" to-layer="1216" to-port="1" />
-		<edge from-layer="1215" from-port="0" to-layer="1220" to-port="1" />
-		<edge from-layer="1216" from-port="2" to-layer="1218" to-port="2" />
-		<edge from-layer="1217" from-port="0" to-layer="1221" to-port="3" />
-		<edge from-layer="1217" from-port="0" to-layer="1218" to-port="3" />
-		<edge from-layer="1218" from-port="4" to-layer="1223" to-port="1" />
-		<edge from-layer="1219" from-port="0" to-layer="1221" to-port="0" />
-		<edge from-layer="1220" from-port="2" to-layer="1221" to-port="2" />
-		<edge from-layer="1220" from-port="2" to-layer="1228" to-port="2" />
-		<edge from-layer="1221" from-port="4" to-layer="1223" to-port="2" />
-		<edge from-layer="1222" from-port="0" to-layer="1223" to-port="3" />
-		<edge from-layer="1223" from-port="4" to-layer="1230" to-port="0" />
-		<edge from-layer="1224" from-port="0" to-layer="1230" to-port="1" />
-		<edge from-layer="1225" from-port="0" to-layer="1228" to-port="0" />
-		<edge from-layer="1226" from-port="0" to-layer="1228" to-port="1" />
-		<edge from-layer="1227" from-port="0" to-layer="1228" to-port="3" />
-		<edge from-layer="1228" from-port="4" to-layer="1230" to-port="2" />
-		<edge from-layer="1229" from-port="0" to-layer="1230" to-port="3" />
-		<edge from-layer="1230" from-port="4" to-layer="1234" to-port="0" />
-		<edge from-layer="1231" from-port="0" to-layer="1232" to-port="0" />
-		<edge from-layer="1232" from-port="1" to-layer="1234" to-port="1" />
-		<edge from-layer="1233" from-port="0" to-layer="1234" to-port="2" />
-		<edge from-layer="1234" from-port="3" to-layer="1235" to-port="0" />
-		<edge from-layer="1235" from-port="2" to-layer="1243" to-port="0" />
-		<edge from-layer="1235" from-port="2" to-layer="1236" to-port="0" />
-		<edge from-layer="1236" from-port="1" to-layer="1242" to-port="0" />
-		<edge from-layer="1237" from-port="0" to-layer="1239" to-port="1" />
-		<edge from-layer="1238" from-port="0" to-layer="1239" to-port="2" />
-		<edge from-layer="1239" from-port="3" to-layer="1240" to-port="0" />
-		<edge from-layer="1240" from-port="4" to-layer="1241" to-port="0" />
-		<edge from-layer="1241" from-port="1" to-layer="1242" to-port="1" />
-		<edge from-layer="1242" from-port="2" to-layer="1243" to-port="1" />
-		<edge from-layer="1243" from-port="2" to-layer="1244" to-port="3" />
-		<edge from-layer="1244" from-port="4" to-layer="1246" to-port="0" />
-		<edge from-layer="1245" from-port="0" to-layer="1246" to-port="1" />
-		<edge from-layer="1246" from-port="2" to-layer="1248" to-port="0" />
-		<edge from-layer="1247" from-port="0" to-layer="1248" to-port="1" />
-		<edge from-layer="1248" from-port="2" to-layer="1257" to-port="0" />
-		<edge from-layer="1249" from-port="0" to-layer="1250" to-port="0" />
-		<edge from-layer="1250" from-port="1" to-layer="1253" to-port="0" />
-		<edge from-layer="1251" from-port="0" to-layer="1252" to-port="0" />
+		<edge from-layer="1201" from-port="4" to-layer="1202" to-port="1" />
+		<edge from-layer="1202" from-port="2" to-layer="1203" to-port="0" />
+		<edge from-layer="1203" from-port="1" to-layer="1205" to-port="0" />
+		<edge from-layer="1204" from-port="0" to-layer="1205" to-port="2" />
+		<edge from-layer="1205" from-port="3" to-layer="1212" to-port="0" />
+		<edge from-layer="1206" from-port="0" to-layer="1209" to-port="1" />
+		<edge from-layer="1207" from-port="0" to-layer="1209" to-port="2" />
+		<edge from-layer="1208" from-port="0" to-layer="1209" to-port="3" />
+		<edge from-layer="1209" from-port="5" to-layer="1211" to-port="0" />
+		<edge from-layer="1210" from-port="0" to-layer="1211" to-port="1" />
+		<edge from-layer="1211" from-port="2" to-layer="1212" to-port="1" />
+		<edge from-layer="1212" from-port="2" to-layer="1221" to-port="2" />
+		<edge from-layer="1212" from-port="2" to-layer="1626" to-port="0" />
+		<edge from-layer="1213" from-port="0" to-layer="1220" to-port="1" />
+		<edge from-layer="1214" from-port="1" to-layer="1217" to-port="0" />
+		<edge from-layer="1215" from-port="0" to-layer="1217" to-port="1" />
+		<edge from-layer="1216" from-port="0" to-layer="1217" to-port="2" />
+		<edge from-layer="1217" from-port="3" to-layer="1218" to-port="0" />
+		<edge from-layer="1218" from-port="2" to-layer="1220" to-port="2" />
+		<edge from-layer="1219" from-port="0" to-layer="1220" to-port="3" />
+		<edge from-layer="1220" from-port="5" to-layer="1221" to-port="3" />
+		<edge from-layer="1221" from-port="4" to-layer="1223" to-port="0" />
+		<edge from-layer="1222" from-port="0" to-layer="1223" to-port="1" />
+		<edge from-layer="1223" from-port="2" to-layer="1225" to-port="0" />
+		<edge from-layer="1224" from-port="0" to-layer="1225" to-port="1" />
+		<edge from-layer="1225" from-port="2" to-layer="1234" to-port="0" />
+		<edge from-layer="1226" from-port="0" to-layer="1227" to-port="0" />
+		<edge from-layer="1227" from-port="1" to-layer="1230" to-port="0" />
+		<edge from-layer="1228" from-port="0" to-layer="1229" to-port="0" />
+		<edge from-layer="1229" from-port="1" to-layer="1230" to-port="1" />
+		<edge from-layer="1230" from-port="2" to-layer="1232" to-port="0" />
+		<edge from-layer="1231" from-port="0" to-layer="1232" to-port="1" />
+		<edge from-layer="1232" from-port="2" to-layer="1233" to-port="0" />
+		<edge from-layer="1233" from-port="1" to-layer="1234" to-port="1" />
+		<edge from-layer="1234" from-port="2" to-layer="1236" to-port="0" />
+		<edge from-layer="1235" from-port="0" to-layer="1236" to-port="1" />
+		<edge from-layer="1236" from-port="2" to-layer="1237" to-port="1" />
+		<edge from-layer="1237" from-port="2" to-layer="1238" to-port="0" />
+		<edge from-layer="1238" from-port="2" to-layer="1396" to-port="1" />
+		<edge from-layer="1238" from-port="2" to-layer="1240" to-port="0" />
+		<edge from-layer="1239" from-port="0" to-layer="1240" to-port="1" />
+		<edge from-layer="1240" from-port="2" to-layer="1271" to-port="0" />
+		<edge from-layer="1240" from-port="2" to-layer="1242" to-port="0" />
+		<edge from-layer="1241" from-port="0" to-layer="1242" to-port="1" />
+		<edge from-layer="1242" from-port="2" to-layer="1244" to-port="0" />
+		<edge from-layer="1243" from-port="0" to-layer="1244" to-port="1" />
+		<edge from-layer="1244" from-port="2" to-layer="1253" to-port="0" />
+		<edge from-layer="1245" from-port="0" to-layer="1246" to-port="0" />
+		<edge from-layer="1246" from-port="1" to-layer="1249" to-port="0" />
+		<edge from-layer="1247" from-port="0" to-layer="1248" to-port="0" />
+		<edge from-layer="1248" from-port="1" to-layer="1249" to-port="1" />
+		<edge from-layer="1249" from-port="2" to-layer="1251" to-port="0" />
+		<edge from-layer="1250" from-port="0" to-layer="1251" to-port="1" />
+		<edge from-layer="1251" from-port="2" to-layer="1252" to-port="0" />
 		<edge from-layer="1252" from-port="1" to-layer="1253" to-port="1" />
-		<edge from-layer="1253" from-port="2" to-layer="1255" to-port="0" />
-		<edge from-layer="1254" from-port="0" to-layer="1255" to-port="1" />
-		<edge from-layer="1255" from-port="2" to-layer="1256" to-port="0" />
-		<edge from-layer="1256" from-port="1" to-layer="1257" to-port="1" />
-		<edge from-layer="1257" from-port="2" to-layer="1260" to-port="0" />
+		<edge from-layer="1253" from-port="2" to-layer="1256" to-port="0" />
+		<edge from-layer="1254" from-port="0" to-layer="1255" to-port="0" />
+		<edge from-layer="1255" from-port="1" to-layer="1256" to-port="1" />
+		<edge from-layer="1256" from-port="2" to-layer="1257" to-port="0" />
+		<edge from-layer="1257" from-port="1" to-layer="1266" to-port="0" />
 		<edge from-layer="1258" from-port="0" to-layer="1259" to-port="0" />
-		<edge from-layer="1259" from-port="1" to-layer="1260" to-port="1" />
-		<edge from-layer="1260" from-port="2" to-layer="1261" to-port="1" />
-		<edge from-layer="1261" from-port="2" to-layer="1262" to-port="0" />
+		<edge from-layer="1259" from-port="1" to-layer="1262" to-port="0" />
+		<edge from-layer="1260" from-port="0" to-layer="1261" to-port="0" />
+		<edge from-layer="1261" from-port="1" to-layer="1262" to-port="1" />
 		<edge from-layer="1262" from-port="2" to-layer="1264" to-port="0" />
-		<edge from-layer="1262" from-port="2" to-layer="1504" to-port="1" />
 		<edge from-layer="1263" from-port="0" to-layer="1264" to-port="1" />
-		<edge from-layer="1264" from-port="2" to-layer="1267" to-port="0" />
-		<edge from-layer="1264" from-port="2" to-layer="1298" to-port="0" />
-		<edge from-layer="1265" from-port="0" to-layer="1266" to-port="0" />
-		<edge from-layer="1266" from-port="1" to-layer="1267" to-port="1" />
-		<edge from-layer="1267" from-port="2" to-layer="1270" to-port="0" />
-		<edge from-layer="1268" from-port="0" to-layer="1269" to-port="0" />
-		<edge from-layer="1269" from-port="1" to-layer="1270" to-port="1" />
-		<edge from-layer="1270" from-port="2" to-layer="1279" to-port="0" />
-		<edge from-layer="1271" from-port="0" to-layer="1272" to-port="0" />
-		<edge from-layer="1272" from-port="1" to-layer="1275" to-port="0" />
-		<edge from-layer="1273" from-port="0" to-layer="1274" to-port="0" />
-		<edge from-layer="1274" from-port="1" to-layer="1275" to-port="1" />
-		<edge from-layer="1275" from-port="2" to-layer="1277" to-port="0" />
-		<edge from-layer="1276" from-port="0" to-layer="1277" to-port="1" />
-		<edge from-layer="1277" from-port="2" to-layer="1278" to-port="0" />
-		<edge from-layer="1278" from-port="1" to-layer="1279" to-port="1" />
-		<edge from-layer="1279" from-port="2" to-layer="1282" to-port="0" />
-		<edge from-layer="1280" from-port="0" to-layer="1281" to-port="0" />
+		<edge from-layer="1264" from-port="2" to-layer="1265" to-port="0" />
+		<edge from-layer="1265" from-port="1" to-layer="1266" to-port="1" />
+		<edge from-layer="1266" from-port="2" to-layer="1269" to-port="0" />
+		<edge from-layer="1267" from-port="0" to-layer="1268" to-port="0" />
+		<edge from-layer="1268" from-port="1" to-layer="1269" to-port="1" />
+		<edge from-layer="1269" from-port="2" to-layer="1395" to-port="0" />
+		<edge from-layer="1270" from-port="0" to-layer="1271" to-port="1" />
+		<edge from-layer="1271" from-port="2" to-layer="1273" to-port="0" />
+		<edge from-layer="1272" from-port="0" to-layer="1273" to-port="1" />
+		<edge from-layer="1273" from-port="2" to-layer="1282" to-port="0" />
+		<edge from-layer="1274" from-port="0" to-layer="1275" to-port="0" />
+		<edge from-layer="1275" from-port="1" to-layer="1278" to-port="0" />
+		<edge from-layer="1276" from-port="0" to-layer="1277" to-port="0" />
+		<edge from-layer="1277" from-port="1" to-layer="1278" to-port="1" />
+		<edge from-layer="1278" from-port="2" to-layer="1280" to-port="0" />
+		<edge from-layer="1279" from-port="0" to-layer="1280" to-port="1" />
+		<edge from-layer="1280" from-port="2" to-layer="1281" to-port="0" />
 		<edge from-layer="1281" from-port="1" to-layer="1282" to-port="1" />
-		<edge from-layer="1282" from-port="2" to-layer="1283" to-port="0" />
-		<edge from-layer="1283" from-port="1" to-layer="1292" to-port="0" />
-		<edge from-layer="1284" from-port="0" to-layer="1285" to-port="0" />
-		<edge from-layer="1285" from-port="1" to-layer="1288" to-port="0" />
-		<edge from-layer="1286" from-port="0" to-layer="1287" to-port="0" />
-		<edge from-layer="1287" from-port="1" to-layer="1288" to-port="1" />
-		<edge from-layer="1288" from-port="2" to-layer="1290" to-port="0" />
-		<edge from-layer="1289" from-port="0" to-layer="1290" to-port="1" />
-		<edge from-layer="1290" from-port="2" to-layer="1291" to-port="0" />
-		<edge from-layer="1291" from-port="1" to-layer="1292" to-port="1" />
-		<edge from-layer="1292" from-port="2" to-layer="1295" to-port="0" />
-		<edge from-layer="1293" from-port="0" to-layer="1294" to-port="0" />
-		<edge from-layer="1294" from-port="1" to-layer="1295" to-port="1" />
-		<edge from-layer="1295" from-port="2" to-layer="1503" to-port="0" />
-		<edge from-layer="1296" from-port="0" to-layer="1297" to-port="0" />
-		<edge from-layer="1297" from-port="1" to-layer="1298" to-port="1" />
-		<edge from-layer="1298" from-port="2" to-layer="1301" to-port="0" />
-		<edge from-layer="1299" from-port="0" to-layer="1300" to-port="0" />
-		<edge from-layer="1300" from-port="1" to-layer="1301" to-port="1" />
-		<edge from-layer="1301" from-port="2" to-layer="1310" to-port="0" />
-		<edge from-layer="1302" from-port="0" to-layer="1303" to-port="0" />
-		<edge from-layer="1303" from-port="1" to-layer="1306" to-port="0" />
-		<edge from-layer="1304" from-port="0" to-layer="1305" to-port="0" />
-		<edge from-layer="1305" from-port="1" to-layer="1306" to-port="1" />
-		<edge from-layer="1306" from-port="2" to-layer="1308" to-port="0" />
-		<edge from-layer="1307" from-port="0" to-layer="1308" to-port="1" />
+		<edge from-layer="1282" from-port="2" to-layer="1284" to-port="0" />
+		<edge from-layer="1283" from-port="0" to-layer="1284" to-port="1" />
+		<edge from-layer="1284" from-port="2" to-layer="1286" to-port="0" />
+		<edge from-layer="1285" from-port="0" to-layer="1286" to-port="1" />
+		<edge from-layer="1286" from-port="2" to-layer="1328" to-port="0" />
+		<edge from-layer="1286" from-port="2" to-layer="1367" to-port="0" />
+		<edge from-layer="1286" from-port="2" to-layer="1290" to-port="0" />
+		<edge from-layer="1287" from-port="0" to-layer="1290" to-port="1" />
+		<edge from-layer="1288" from-port="0" to-layer="1290" to-port="2" />
+		<edge from-layer="1289" from-port="0" to-layer="1290" to-port="3" />
+		<edge from-layer="1290" from-port="5" to-layer="1292" to-port="0" />
+		<edge from-layer="1291" from-port="0" to-layer="1292" to-port="1" />
+		<edge from-layer="1292" from-port="2" to-layer="1296" to-port="0" />
+		<edge from-layer="1292" from-port="2" to-layer="1314" to-port="0" />
+		<edge from-layer="1293" from-port="0" to-layer="1296" to-port="1" />
+		<edge from-layer="1294" from-port="0" to-layer="1296" to-port="2" />
+		<edge from-layer="1295" from-port="0" to-layer="1296" to-port="3" />
+		<edge from-layer="1296" from-port="5" to-layer="1301" to-port="0" />
+		<edge from-layer="1296" from-port="5" to-layer="1297" to-port="0" />
+		<edge from-layer="1296" from-port="5" to-layer="1307" to-port="0" />
+		<edge from-layer="1297" from-port="2" to-layer="1310" to-port="0" />
+		<edge from-layer="1298" from-port="0" to-layer="1301" to-port="1" />
+		<edge from-layer="1299" from-port="0" to-layer="1301" to-port="2" />
+		<edge from-layer="1300" from-port="0" to-layer="1301" to-port="3" />
+		<edge from-layer="1301" from-port="5" to-layer="1303" to-port="0" />
+		<edge from-layer="1302" from-port="0" to-layer="1303" to-port="1" />
+		<edge from-layer="1303" from-port="2" to-layer="1308" to-port="0" />
+		<edge from-layer="1304" from-port="0" to-layer="1307" to-port="1" />
+		<edge from-layer="1305" from-port="0" to-layer="1307" to-port="2" />
+		<edge from-layer="1306" from-port="0" to-layer="1307" to-port="3" />
+		<edge from-layer="1307" from-port="5" to-layer="1308" to-port="1" />
 		<edge from-layer="1308" from-port="2" to-layer="1309" to-port="0" />
-		<edge from-layer="1309" from-port="1" to-layer="1310" to-port="1" />
-		<edge from-layer="1310" from-port="2" to-layer="1313" to-port="0" />
-		<edge from-layer="1311" from-port="0" to-layer="1312" to-port="0" />
-		<edge from-layer="1312" from-port="1" to-layer="1313" to-port="1" />
-		<edge from-layer="1313" from-port="2" to-layer="1315" to-port="0" />
-		<edge from-layer="1314" from-port="0" to-layer="1315" to-port="1" />
-		<edge from-layer="1315" from-port="2" to-layer="1319" to-port="0" />
-		<edge from-layer="1315" from-port="2" to-layer="1443" to-port="0" />
-		<edge from-layer="1315" from-port="2" to-layer="1332" to-port="0" />
-		<edge from-layer="1316" from-port="0" to-layer="1319" to-port="1" />
-		<edge from-layer="1317" from-port="0" to-layer="1319" to-port="2" />
-		<edge from-layer="1318" from-port="0" to-layer="1319" to-port="3" />
-		<edge from-layer="1319" from-port="4" to-layer="1321" to-port="0" />
-		<edge from-layer="1320" from-port="0" to-layer="1321" to-port="1" />
-		<edge from-layer="1321" from-port="2" to-layer="1325" to-port="0" />
-		<edge from-layer="1321" from-port="2" to-layer="1394" to-port="0" />
-		<edge from-layer="1322" from-port="0" to-layer="1325" to-port="1" />
-		<edge from-layer="1323" from-port="0" to-layer="1325" to-port="2" />
-		<edge from-layer="1324" from-port="0" to-layer="1325" to-port="3" />
-		<edge from-layer="1325" from-port="4" to-layer="1379" to-port="0" />
-		<edge from-layer="1325" from-port="4" to-layer="1373" to-port="0" />
-		<edge from-layer="1325" from-port="4" to-layer="1362" to-port="0" />
-		<edge from-layer="1325" from-port="4" to-layer="1359" to-port="0" />
-		<edge from-layer="1326" from-port="0" to-layer="1327" to-port="0" />
-		<edge from-layer="1327" from-port="1" to-layer="1355" to-port="0" />
-		<edge from-layer="1328" from-port="0" to-layer="1355" to-port="1" />
-		<edge from-layer="1329" from-port="0" to-layer="1332" to-port="1" />
-		<edge from-layer="1330" from-port="0" to-layer="1332" to-port="2" />
-		<edge from-layer="1331" from-port="0" to-layer="1332" to-port="3" />
-		<edge from-layer="1332" from-port="4" to-layer="1334" to-port="0" />
-		<edge from-layer="1333" from-port="0" to-layer="1334" to-port="1" />
-		<edge from-layer="1334" from-port="2" to-layer="1335" to-port="0" />
-		<edge from-layer="1334" from-port="2" to-layer="1399" to-port="0" />
-		<edge from-layer="1334" from-port="2" to-layer="1427" to-port="0" />
-		<edge from-layer="1335" from-port="1" to-layer="1337" to-port="0" />
-		<edge from-layer="1336" from-port="0" to-layer="1337" to-port="2" />
-		<edge from-layer="1337" from-port="3" to-layer="1351" to-port="0" />
-		<edge from-layer="1338" from-port="0" to-layer="1339" to-port="0" />
-		<edge from-layer="1339" from-port="1" to-layer="1344" to-port="0" />
-		<edge from-layer="1340" from-port="0" to-layer="1343" to-port="1" />
-		<edge from-layer="1341" from-port="0" to-layer="1343" to-port="2" />
-		<edge from-layer="1342" from-port="0" to-layer="1343" to-port="3" />
-		<edge from-layer="1343" from-port="4" to-layer="1344" to-port="1" />
-		<edge from-layer="1344" from-port="2" to-layer="1345" to-port="0" />
-		<edge from-layer="1345" from-port="1" to-layer="1347" to-port="0" />
-		<edge from-layer="1346" from-port="0" to-layer="1347" to-port="2" />
-		<edge from-layer="1347" from-port="3" to-layer="1429" to-port="0" />
-		<edge from-layer="1347" from-port="3" to-layer="1348" to-port="0" />
-		<edge from-layer="1348" from-port="1" to-layer="1350" to-port="0" />
-		<edge from-layer="1349" from-port="0" to-layer="1350" to-port="2" />
-		<edge from-layer="1350" from-port="3" to-layer="1351" to-port="1" />
-		<edge from-layer="1351" from-port="2" to-layer="1353" to-port="0" />
-		<edge from-layer="1352" from-port="0" to-layer="1353" to-port="1" />
-		<edge from-layer="1353" from-port="2" to-layer="1355" to-port="2" />
-		<edge from-layer="1353" from-port="2" to-layer="1385" to-port="2" />
-		<edge from-layer="1354" from-port="0" to-layer="1355" to-port="3" />
-		<edge from-layer="1355" from-port="4" to-layer="1357" to-port="0" />
-		<edge from-layer="1356" from-port="0" to-layer="1357" to-port="2" />
-		<edge from-layer="1357" from-port="3" to-layer="1358" to-port="0" />
-		<edge from-layer="1358" from-port="2" to-layer="1400" to-port="1" />
-		<edge from-layer="1358" from-port="2" to-layer="1359" to-port="1" />
-		<edge from-layer="1359" from-port="2" to-layer="1390" to-port="0" />
-		<edge from-layer="1360" from-port="0" to-layer="1370" to-port="0" />
-		<edge from-layer="1361" from-port="0" to-layer="1370" to-port="1" />
-		<edge from-layer="1362" from-port="1" to-layer="1364" to-port="0" />
-		<edge from-layer="1363" from-port="0" to-layer="1364" to-port="2" />
-		<edge from-layer="1364" from-port="3" to-layer="1365" to-port="0" />
-		<edge from-layer="1365" from-port="2" to-layer="1366" to-port="0" />
-		<edge from-layer="1366" from-port="1" to-layer="1368" to-port="0" />
-		<edge from-layer="1367" from-port="0" to-layer="1368" to-port="1" />
-		<edge from-layer="1368" from-port="2" to-layer="1370" to-port="2" />
-		<edge from-layer="1369" from-port="0" to-layer="1370" to-port="3" />
-		<edge from-layer="1370" from-port="4" to-layer="1373" to-port="1" />
-		<edge from-layer="1370" from-port="4" to-layer="1379" to-port="2" />
-		<edge from-layer="1371" from-port="0" to-layer="1373" to-port="2" />
-		<edge from-layer="1372" from-port="0" to-layer="1373" to-port="3" />
-		<edge from-layer="1373" from-port="4" to-layer="1376" to-port="0" />
-		<edge from-layer="1374" from-port="0" to-layer="1375" to-port="0" />
-		<edge from-layer="1375" from-port="1" to-layer="1376" to-port="1" />
-		<edge from-layer="1376" from-port="2" to-layer="1380" to-port="0" />
-		<edge from-layer="1377" from-port="0" to-layer="1379" to-port="1" />
-		<edge from-layer="1378" from-port="0" to-layer="1379" to-port="3" />
-		<edge from-layer="1379" from-port="4" to-layer="1380" to-port="1" />
-		<edge from-layer="1380" from-port="2" to-layer="1389" to-port="0" />
-		<edge from-layer="1381" from-port="0" to-layer="1382" to-port="0" />
-		<edge from-layer="1382" from-port="1" to-layer="1385" to-port="0" />
-		<edge from-layer="1383" from-port="0" to-layer="1385" to-port="1" />
-		<edge from-layer="1384" from-port="0" to-layer="1385" to-port="3" />
-		<edge from-layer="1385" from-port="4" to-layer="1387" to-port="0" />
-		<edge from-layer="1386" from-port="0" to-layer="1387" to-port="2" />
-		<edge from-layer="1387" from-port="3" to-layer="1388" to-port="0" />
-		<edge from-layer="1388" from-port="2" to-layer="1389" to-port="1" />
-		<edge from-layer="1388" from-port="2" to-layer="1422" to-port="1" />
-		<edge from-layer="1389" from-port="2" to-layer="1390" to-port="1" />
-		<edge from-layer="1390" from-port="2" to-layer="1395" to-port="0" />
-		<edge from-layer="1391" from-port="0" to-layer="1394" to-port="1" />
-		<edge from-layer="1392" from-port="0" to-layer="1394" to-port="2" />
-		<edge from-layer="1393" from-port="0" to-layer="1394" to-port="3" />
-		<edge from-layer="1394" from-port="4" to-layer="1395" to-port="1" />
-		<edge from-layer="1395" from-port="2" to-layer="1453" to-port="0" />
-		<edge from-layer="1395" from-port="2" to-layer="1486" to-port="0" />
-		<edge from-layer="1396" from-port="0" to-layer="1399" to-port="1" />
-		<edge from-layer="1397" from-port="0" to-layer="1399" to-port="2" />
-		<edge from-layer="1398" from-port="0" to-layer="1399" to-port="3" />
-		<edge from-layer="1399" from-port="4" to-layer="1420" to-port="0" />
-		<edge from-layer="1399" from-port="4" to-layer="1403" to-port="0" />
-		<edge from-layer="1399" from-port="4" to-layer="1400" to-port="0" />
-		<edge from-layer="1399" from-port="4" to-layer="1414" to-port="0" />
-		<edge from-layer="1400" from-port="2" to-layer="1423" to-port="0" />
-		<edge from-layer="1401" from-port="0" to-layer="1411" to-port="0" />
-		<edge from-layer="1402" from-port="0" to-layer="1411" to-port="1" />
-		<edge from-layer="1403" from-port="1" to-layer="1405" to-port="0" />
-		<edge from-layer="1404" from-port="0" to-layer="1405" to-port="2" />
-		<edge from-layer="1405" from-port="3" to-layer="1406" to-port="0" />
-		<edge from-layer="1406" from-port="2" to-layer="1407" to-port="0" />
-		<edge from-layer="1407" from-port="1" to-layer="1409" to-port="0" />
+		<edge from-layer="1309" from-port="2" to-layer="1310" to-port="1" />
+		<edge from-layer="1310" from-port="2" to-layer="1315" to-port="0" />
+		<edge from-layer="1311" from-port="0" to-layer="1314" to-port="1" />
+		<edge from-layer="1312" from-port="0" to-layer="1314" to-port="2" />
+		<edge from-layer="1313" from-port="0" to-layer="1314" to-port="3" />
+		<edge from-layer="1314" from-port="5" to-layer="1315" to-port="1" />
+		<edge from-layer="1315" from-port="2" to-layer="1379" to-port="0" />
+		<edge from-layer="1316" from-port="0" to-layer="1321" to-port="0" />
+		<edge from-layer="1317" from-port="0" to-layer="1320" to-port="1" />
+		<edge from-layer="1318" from-port="0" to-layer="1320" to-port="2" />
+		<edge from-layer="1319" from-port="0" to-layer="1320" to-port="3" />
+		<edge from-layer="1320" from-port="4" to-layer="1321" to-port="1" />
+		<edge from-layer="1321" from-port="2" to-layer="1322" to-port="0" />
+		<edge from-layer="1322" from-port="1" to-layer="1324" to-port="0" />
+		<edge from-layer="1323" from-port="0" to-layer="1324" to-port="2" />
+		<edge from-layer="1324" from-port="3" to-layer="1354" to-port="0" />
+		<edge from-layer="1324" from-port="3" to-layer="1372" to-port="0" />
+		<edge from-layer="1325" from-port="0" to-layer="1328" to-port="1" />
+		<edge from-layer="1326" from-port="0" to-layer="1328" to-port="2" />
+		<edge from-layer="1327" from-port="0" to-layer="1328" to-port="3" />
+		<edge from-layer="1328" from-port="5" to-layer="1330" to-port="0" />
+		<edge from-layer="1329" from-port="0" to-layer="1330" to-port="1" />
+		<edge from-layer="1330" from-port="2" to-layer="1334" to-port="0" />
+		<edge from-layer="1330" from-port="2" to-layer="1352" to-port="0" />
+		<edge from-layer="1331" from-port="0" to-layer="1334" to-port="1" />
+		<edge from-layer="1332" from-port="0" to-layer="1334" to-port="2" />
+		<edge from-layer="1333" from-port="0" to-layer="1334" to-port="3" />
+		<edge from-layer="1334" from-port="5" to-layer="1335" to-port="0" />
+		<edge from-layer="1334" from-port="5" to-layer="1339" to-port="0" />
+		<edge from-layer="1334" from-port="5" to-layer="1345" to-port="0" />
+		<edge from-layer="1335" from-port="2" to-layer="1348" to-port="0" />
+		<edge from-layer="1336" from-port="0" to-layer="1339" to-port="1" />
+		<edge from-layer="1337" from-port="0" to-layer="1339" to-port="2" />
+		<edge from-layer="1338" from-port="0" to-layer="1339" to-port="3" />
+		<edge from-layer="1339" from-port="5" to-layer="1341" to-port="0" />
+		<edge from-layer="1340" from-port="0" to-layer="1341" to-port="1" />
+		<edge from-layer="1341" from-port="2" to-layer="1346" to-port="0" />
+		<edge from-layer="1342" from-port="0" to-layer="1345" to-port="1" />
+		<edge from-layer="1343" from-port="0" to-layer="1345" to-port="2" />
+		<edge from-layer="1344" from-port="0" to-layer="1345" to-port="3" />
+		<edge from-layer="1345" from-port="5" to-layer="1346" to-port="1" />
+		<edge from-layer="1346" from-port="2" to-layer="1347" to-port="0" />
+		<edge from-layer="1347" from-port="2" to-layer="1348" to-port="1" />
+		<edge from-layer="1348" from-port="2" to-layer="1353" to-port="0" />
+		<edge from-layer="1349" from-port="0" to-layer="1352" to-port="1" />
+		<edge from-layer="1350" from-port="0" to-layer="1352" to-port="2" />
+		<edge from-layer="1351" from-port="0" to-layer="1352" to-port="3" />
+		<edge from-layer="1352" from-port="5" to-layer="1353" to-port="1" />
+		<edge from-layer="1353" from-port="2" to-layer="1354" to-port="1" />
+		<edge from-layer="1354" from-port="2" to-layer="1379" to-port="1" />
+		<edge from-layer="1354" from-port="2" to-layer="1625" to-port="0" />
+		<edge from-layer="1355" from-port="0" to-layer="1360" to-port="0" />
+		<edge from-layer="1356" from-port="0" to-layer="1359" to-port="1" />
+		<edge from-layer="1357" from-port="0" to-layer="1359" to-port="2" />
+		<edge from-layer="1358" from-port="0" to-layer="1359" to-port="3" />
+		<edge from-layer="1359" from-port="4" to-layer="1360" to-port="1" />
+		<edge from-layer="1360" from-port="2" to-layer="1361" to-port="0" />
+		<edge from-layer="1361" from-port="1" to-layer="1363" to-port="0" />
+		<edge from-layer="1362" from-port="0" to-layer="1363" to-port="2" />
+		<edge from-layer="1363" from-port="3" to-layer="1370" to-port="0" />
+		<edge from-layer="1364" from-port="0" to-layer="1367" to-port="1" />
+		<edge from-layer="1365" from-port="0" to-layer="1367" to-port="2" />
+		<edge from-layer="1366" from-port="0" to-layer="1367" to-port="3" />
+		<edge from-layer="1367" from-port="5" to-layer="1369" to-port="0" />
+		<edge from-layer="1368" from-port="0" to-layer="1369" to-port="1" />
+		<edge from-layer="1369" from-port="2" to-layer="1370" to-port="1" />
+		<edge from-layer="1370" from-port="2" to-layer="1379" to-port="2" />
+		<edge from-layer="1370" from-port="2" to-layer="1624" to-port="0" />
+		<edge from-layer="1371" from-port="0" to-layer="1378" to-port="1" />
+		<edge from-layer="1372" from-port="1" to-layer="1375" to-port="0" />
+		<edge from-layer="1373" from-port="0" to-layer="1375" to-port="1" />
+		<edge from-layer="1374" from-port="0" to-layer="1375" to-port="2" />
+		<edge from-layer="1375" from-port="3" to-layer="1376" to-port="0" />
+		<edge from-layer="1376" from-port="2" to-layer="1378" to-port="2" />
+		<edge from-layer="1377" from-port="0" to-layer="1378" to-port="3" />
+		<edge from-layer="1378" from-port="5" to-layer="1379" to-port="3" />
+		<edge from-layer="1379" from-port="4" to-layer="1381" to-port="0" />
+		<edge from-layer="1380" from-port="0" to-layer="1381" to-port="1" />
+		<edge from-layer="1381" from-port="2" to-layer="1383" to-port="0" />
+		<edge from-layer="1382" from-port="0" to-layer="1383" to-port="1" />
+		<edge from-layer="1383" from-port="2" to-layer="1392" to-port="0" />
+		<edge from-layer="1384" from-port="0" to-layer="1385" to-port="0" />
+		<edge from-layer="1385" from-port="1" to-layer="1388" to-port="0" />
+		<edge from-layer="1386" from-port="0" to-layer="1387" to-port="0" />
+		<edge from-layer="1387" from-port="1" to-layer="1388" to-port="1" />
+		<edge from-layer="1388" from-port="2" to-layer="1390" to-port="0" />
+		<edge from-layer="1389" from-port="0" to-layer="1390" to-port="1" />
+		<edge from-layer="1390" from-port="2" to-layer="1391" to-port="0" />
+		<edge from-layer="1391" from-port="1" to-layer="1392" to-port="1" />
+		<edge from-layer="1392" from-port="2" to-layer="1394" to-port="0" />
+		<edge from-layer="1393" from-port="0" to-layer="1394" to-port="1" />
+		<edge from-layer="1394" from-port="2" to-layer="1395" to-port="1" />
+		<edge from-layer="1395" from-port="2" to-layer="1396" to-port="0" />
+		<edge from-layer="1396" from-port="2" to-layer="1554" to-port="1" />
+		<edge from-layer="1396" from-port="2" to-layer="1398" to-port="0" />
+		<edge from-layer="1397" from-port="0" to-layer="1398" to-port="1" />
+		<edge from-layer="1398" from-port="2" to-layer="1429" to-port="0" />
+		<edge from-layer="1398" from-port="2" to-layer="1400" to-port="0" />
+		<edge from-layer="1399" from-port="0" to-layer="1400" to-port="1" />
+		<edge from-layer="1400" from-port="2" to-layer="1402" to-port="0" />
+		<edge from-layer="1401" from-port="0" to-layer="1402" to-port="1" />
+		<edge from-layer="1402" from-port="2" to-layer="1411" to-port="0" />
+		<edge from-layer="1403" from-port="0" to-layer="1404" to-port="0" />
+		<edge from-layer="1404" from-port="1" to-layer="1407" to-port="0" />
+		<edge from-layer="1405" from-port="0" to-layer="1406" to-port="0" />
+		<edge from-layer="1406" from-port="1" to-layer="1407" to-port="1" />
+		<edge from-layer="1407" from-port="2" to-layer="1409" to-port="0" />
 		<edge from-layer="1408" from-port="0" to-layer="1409" to-port="1" />
-		<edge from-layer="1409" from-port="2" to-layer="1411" to-port="2" />
-		<edge from-layer="1410" from-port="0" to-layer="1411" to-port="3" />
-		<edge from-layer="1411" from-port="4" to-layer="1420" to-port="2" />
-		<edge from-layer="1411" from-port="4" to-layer="1414" to-port="1" />
-		<edge from-layer="1412" from-port="0" to-layer="1414" to-port="2" />
-		<edge from-layer="1413" from-port="0" to-layer="1414" to-port="3" />
-		<edge from-layer="1414" from-port="4" to-layer="1417" to-port="0" />
-		<edge from-layer="1415" from-port="0" to-layer="1416" to-port="0" />
-		<edge from-layer="1416" from-port="1" to-layer="1417" to-port="1" />
-		<edge from-layer="1417" from-port="2" to-layer="1421" to-port="0" />
-		<edge from-layer="1418" from-port="0" to-layer="1420" to-port="1" />
-		<edge from-layer="1419" from-port="0" to-layer="1420" to-port="3" />
-		<edge from-layer="1420" from-port="4" to-layer="1421" to-port="1" />
-		<edge from-layer="1421" from-port="2" to-layer="1422" to-port="0" />
-		<edge from-layer="1422" from-port="2" to-layer="1423" to-port="1" />
-		<edge from-layer="1423" from-port="2" to-layer="1428" to-port="0" />
-		<edge from-layer="1424" from-port="0" to-layer="1427" to-port="1" />
-		<edge from-layer="1425" from-port="0" to-layer="1427" to-port="2" />
-		<edge from-layer="1426" from-port="0" to-layer="1427" to-port="3" />
-		<edge from-layer="1427" from-port="4" to-layer="1428" to-port="1" />
-		<edge from-layer="1428" from-port="2" to-layer="1429" to-port="1" />
-		<edge from-layer="1429" from-port="2" to-layer="1486" to-port="1" />
-		<edge from-layer="1429" from-port="2" to-layer="2344" to-port="0" />
-		<edge from-layer="1429" from-port="2" to-layer="1450" to-port="0" />
-		<edge from-layer="1430" from-port="0" to-layer="1431" to-port="0" />
-		<edge from-layer="1431" from-port="1" to-layer="1436" to-port="0" />
-		<edge from-layer="1432" from-port="0" to-layer="1435" to-port="1" />
-		<edge from-layer="1433" from-port="0" to-layer="1435" to-port="2" />
-		<edge from-layer="1434" from-port="0" to-layer="1435" to-port="3" />
-		<edge from-layer="1435" from-port="4" to-layer="1436" to-port="1" />
-		<edge from-layer="1436" from-port="2" to-layer="1437" to-port="0" />
-		<edge from-layer="1437" from-port="1" to-layer="1439" to-port="0" />
-		<edge from-layer="1438" from-port="0" to-layer="1439" to-port="2" />
-		<edge from-layer="1439" from-port="3" to-layer="1446" to-port="0" />
-		<edge from-layer="1440" from-port="0" to-layer="1443" to-port="1" />
-		<edge from-layer="1441" from-port="0" to-layer="1443" to-port="2" />
-		<edge from-layer="1442" from-port="0" to-layer="1443" to-port="3" />
-		<edge from-layer="1443" from-port="4" to-layer="1445" to-port="0" />
-		<edge from-layer="1444" from-port="0" to-layer="1445" to-port="1" />
-		<edge from-layer="1445" from-port="2" to-layer="1446" to-port="1" />
-		<edge from-layer="1446" from-port="2" to-layer="2343" to-port="0" />
-		<edge from-layer="1446" from-port="2" to-layer="1486" to-port="2" />
-		<edge from-layer="1447" from-port="0" to-layer="1465" to-port="0" />
-		<edge from-layer="1448" from-port="0" to-layer="1460" to-port="0" />
-		<edge from-layer="1449" from-port="0" to-layer="1460" to-port="1" />
-		<edge from-layer="1449" from-port="0" to-layer="1463" to-port="1" />
-		<edge from-layer="1450" from-port="1" to-layer="1452" to-port="0" />
-		<edge from-layer="1451" from-port="0" to-layer="1452" to-port="2" />
-		<edge from-layer="1452" from-port="3" to-layer="1456" to-port="0" />
-		<edge from-layer="1452" from-port="3" to-layer="1462" to-port="0" />
-		<edge from-layer="1453" from-port="1" to-layer="1481" to-port="0" />
-		<edge from-layer="1453" from-port="1" to-layer="1455" to-port="0" />
-		<edge from-layer="1454" from-port="0" to-layer="1455" to-port="2" />
-		<edge from-layer="1455" from-port="3" to-layer="1456" to-port="1" />
-		<edge from-layer="1456" from-port="2" to-layer="1458" to-port="0" />
-		<edge from-layer="1457" from-port="0" to-layer="1462" to-port="1" />
-		<edge from-layer="1457" from-port="0" to-layer="1458" to-port="1" />
-		<edge from-layer="1458" from-port="2" to-layer="1460" to-port="2" />
-		<edge from-layer="1459" from-port="0" to-layer="1463" to-port="3" />
-		<edge from-layer="1459" from-port="0" to-layer="1460" to-port="3" />
-		<edge from-layer="1460" from-port="4" to-layer="1465" to-port="1" />
-		<edge from-layer="1461" from-port="0" to-layer="1463" to-port="0" />
-		<edge from-layer="1462" from-port="2" to-layer="1470" to-port="2" />
-		<edge from-layer="1462" from-port="2" to-layer="1463" to-port="2" />
-		<edge from-layer="1463" from-port="4" to-layer="1465" to-port="2" />
+		<edge from-layer="1409" from-port="2" to-layer="1410" to-port="0" />
+		<edge from-layer="1410" from-port="1" to-layer="1411" to-port="1" />
+		<edge from-layer="1411" from-port="2" to-layer="1414" to-port="0" />
+		<edge from-layer="1412" from-port="0" to-layer="1413" to-port="0" />
+		<edge from-layer="1413" from-port="1" to-layer="1414" to-port="1" />
+		<edge from-layer="1414" from-port="2" to-layer="1415" to-port="0" />
+		<edge from-layer="1415" from-port="1" to-layer="1424" to-port="0" />
+		<edge from-layer="1416" from-port="0" to-layer="1417" to-port="0" />
+		<edge from-layer="1417" from-port="1" to-layer="1420" to-port="0" />
+		<edge from-layer="1418" from-port="0" to-layer="1419" to-port="0" />
+		<edge from-layer="1419" from-port="1" to-layer="1420" to-port="1" />
+		<edge from-layer="1420" from-port="2" to-layer="1422" to-port="0" />
+		<edge from-layer="1421" from-port="0" to-layer="1422" to-port="1" />
+		<edge from-layer="1422" from-port="2" to-layer="1423" to-port="0" />
+		<edge from-layer="1423" from-port="1" to-layer="1424" to-port="1" />
+		<edge from-layer="1424" from-port="2" to-layer="1427" to-port="0" />
+		<edge from-layer="1425" from-port="0" to-layer="1426" to-port="0" />
+		<edge from-layer="1426" from-port="1" to-layer="1427" to-port="1" />
+		<edge from-layer="1427" from-port="2" to-layer="1553" to-port="0" />
+		<edge from-layer="1428" from-port="0" to-layer="1429" to-port="1" />
+		<edge from-layer="1429" from-port="2" to-layer="1431" to-port="0" />
+		<edge from-layer="1430" from-port="0" to-layer="1431" to-port="1" />
+		<edge from-layer="1431" from-port="2" to-layer="1440" to-port="0" />
+		<edge from-layer="1432" from-port="0" to-layer="1433" to-port="0" />
+		<edge from-layer="1433" from-port="1" to-layer="1436" to-port="0" />
+		<edge from-layer="1434" from-port="0" to-layer="1435" to-port="0" />
+		<edge from-layer="1435" from-port="1" to-layer="1436" to-port="1" />
+		<edge from-layer="1436" from-port="2" to-layer="1438" to-port="0" />
+		<edge from-layer="1437" from-port="0" to-layer="1438" to-port="1" />
+		<edge from-layer="1438" from-port="2" to-layer="1439" to-port="0" />
+		<edge from-layer="1439" from-port="1" to-layer="1440" to-port="1" />
+		<edge from-layer="1440" from-port="2" to-layer="1442" to-port="0" />
+		<edge from-layer="1441" from-port="0" to-layer="1442" to-port="1" />
+		<edge from-layer="1442" from-port="2" to-layer="1444" to-port="0" />
+		<edge from-layer="1443" from-port="0" to-layer="1444" to-port="1" />
+		<edge from-layer="1444" from-port="2" to-layer="1448" to-port="0" />
+		<edge from-layer="1444" from-port="2" to-layer="1486" to-port="0" />
+		<edge from-layer="1444" from-port="2" to-layer="1525" to-port="0" />
+		<edge from-layer="1445" from-port="0" to-layer="1448" to-port="1" />
+		<edge from-layer="1446" from-port="0" to-layer="1448" to-port="2" />
+		<edge from-layer="1447" from-port="0" to-layer="1448" to-port="3" />
+		<edge from-layer="1448" from-port="5" to-layer="1450" to-port="0" />
+		<edge from-layer="1449" from-port="0" to-layer="1450" to-port="1" />
+		<edge from-layer="1450" from-port="2" to-layer="1454" to-port="0" />
+		<edge from-layer="1450" from-port="2" to-layer="1472" to-port="0" />
+		<edge from-layer="1451" from-port="0" to-layer="1454" to-port="1" />
+		<edge from-layer="1452" from-port="0" to-layer="1454" to-port="2" />
+		<edge from-layer="1453" from-port="0" to-layer="1454" to-port="3" />
+		<edge from-layer="1454" from-port="5" to-layer="1459" to-port="0" />
+		<edge from-layer="1454" from-port="5" to-layer="1465" to-port="0" />
+		<edge from-layer="1454" from-port="5" to-layer="1455" to-port="0" />
+		<edge from-layer="1455" from-port="2" to-layer="1468" to-port="0" />
+		<edge from-layer="1456" from-port="0" to-layer="1459" to-port="1" />
+		<edge from-layer="1457" from-port="0" to-layer="1459" to-port="2" />
+		<edge from-layer="1458" from-port="0" to-layer="1459" to-port="3" />
+		<edge from-layer="1459" from-port="5" to-layer="1461" to-port="0" />
+		<edge from-layer="1460" from-port="0" to-layer="1461" to-port="1" />
+		<edge from-layer="1461" from-port="2" to-layer="1466" to-port="0" />
+		<edge from-layer="1462" from-port="0" to-layer="1465" to-port="1" />
+		<edge from-layer="1463" from-port="0" to-layer="1465" to-port="2" />
 		<edge from-layer="1464" from-port="0" to-layer="1465" to-port="3" />
-		<edge from-layer="1465" from-port="4" to-layer="1472" to-port="0" />
-		<edge from-layer="1466" from-port="0" to-layer="1472" to-port="1" />
-		<edge from-layer="1467" from-port="0" to-layer="1470" to-port="0" />
-		<edge from-layer="1468" from-port="0" to-layer="1470" to-port="1" />
-		<edge from-layer="1469" from-port="0" to-layer="1470" to-port="3" />
-		<edge from-layer="1470" from-port="4" to-layer="1472" to-port="2" />
+		<edge from-layer="1465" from-port="5" to-layer="1466" to-port="1" />
+		<edge from-layer="1466" from-port="2" to-layer="1467" to-port="0" />
+		<edge from-layer="1467" from-port="2" to-layer="1468" to-port="1" />
+		<edge from-layer="1468" from-port="2" to-layer="1473" to-port="0" />
+		<edge from-layer="1469" from-port="0" to-layer="1472" to-port="1" />
+		<edge from-layer="1470" from-port="0" to-layer="1472" to-port="2" />
 		<edge from-layer="1471" from-port="0" to-layer="1472" to-port="3" />
-		<edge from-layer="1472" from-port="4" to-layer="1476" to-port="0" />
-		<edge from-layer="1473" from-port="0" to-layer="1474" to-port="0" />
-		<edge from-layer="1474" from-port="1" to-layer="1476" to-port="1" />
-		<edge from-layer="1475" from-port="0" to-layer="1476" to-port="2" />
-		<edge from-layer="1476" from-port="3" to-layer="1477" to-port="0" />
-		<edge from-layer="1477" from-port="2" to-layer="1478" to-port="0" />
-		<edge from-layer="1477" from-port="2" to-layer="1485" to-port="0" />
-		<edge from-layer="1478" from-port="1" to-layer="1484" to-port="0" />
-		<edge from-layer="1479" from-port="0" to-layer="1481" to-port="1" />
-		<edge from-layer="1480" from-port="0" to-layer="1481" to-port="2" />
-		<edge from-layer="1481" from-port="3" to-layer="1482" to-port="0" />
-		<edge from-layer="1482" from-port="4" to-layer="1483" to-port="0" />
-		<edge from-layer="1483" from-port="1" to-layer="1484" to-port="1" />
-		<edge from-layer="1484" from-port="2" to-layer="1485" to-port="1" />
-		<edge from-layer="1485" from-port="2" to-layer="1486" to-port="3" />
-		<edge from-layer="1486" from-port="4" to-layer="1488" to-port="0" />
+		<edge from-layer="1472" from-port="5" to-layer="1473" to-port="1" />
+		<edge from-layer="1473" from-port="2" to-layer="1537" to-port="0" />
+		<edge from-layer="1474" from-port="0" to-layer="1479" to-port="0" />
+		<edge from-layer="1475" from-port="0" to-layer="1478" to-port="1" />
+		<edge from-layer="1476" from-port="0" to-layer="1478" to-port="2" />
+		<edge from-layer="1477" from-port="0" to-layer="1478" to-port="3" />
+		<edge from-layer="1478" from-port="4" to-layer="1479" to-port="1" />
+		<edge from-layer="1479" from-port="2" to-layer="1480" to-port="0" />
+		<edge from-layer="1480" from-port="1" to-layer="1482" to-port="0" />
+		<edge from-layer="1481" from-port="0" to-layer="1482" to-port="2" />
+		<edge from-layer="1482" from-port="3" to-layer="1512" to-port="0" />
+		<edge from-layer="1482" from-port="3" to-layer="1530" to-port="0" />
+		<edge from-layer="1483" from-port="0" to-layer="1486" to-port="1" />
+		<edge from-layer="1484" from-port="0" to-layer="1486" to-port="2" />
+		<edge from-layer="1485" from-port="0" to-layer="1486" to-port="3" />
+		<edge from-layer="1486" from-port="5" to-layer="1488" to-port="0" />
 		<edge from-layer="1487" from-port="0" to-layer="1488" to-port="1" />
-		<edge from-layer="1488" from-port="2" to-layer="1490" to-port="0" />
-		<edge from-layer="1489" from-port="0" to-layer="1490" to-port="1" />
-		<edge from-layer="1490" from-port="2" to-layer="1499" to-port="0" />
-		<edge from-layer="1491" from-port="0" to-layer="1492" to-port="0" />
-		<edge from-layer="1492" from-port="1" to-layer="1495" to-port="0" />
-		<edge from-layer="1493" from-port="0" to-layer="1494" to-port="0" />
-		<edge from-layer="1494" from-port="1" to-layer="1495" to-port="1" />
-		<edge from-layer="1495" from-port="2" to-layer="1497" to-port="0" />
-		<edge from-layer="1496" from-port="0" to-layer="1497" to-port="1" />
-		<edge from-layer="1497" from-port="2" to-layer="1498" to-port="0" />
-		<edge from-layer="1498" from-port="1" to-layer="1499" to-port="1" />
-		<edge from-layer="1499" from-port="2" to-layer="1502" to-port="0" />
-		<edge from-layer="1500" from-port="0" to-layer="1501" to-port="0" />
-		<edge from-layer="1501" from-port="1" to-layer="1502" to-port="1" />
-		<edge from-layer="1502" from-port="2" to-layer="1503" to-port="1" />
-		<edge from-layer="1503" from-port="2" to-layer="1504" to-port="0" />
-		<edge from-layer="1504" from-port="2" to-layer="1506" to-port="0" />
-		<edge from-layer="1504" from-port="2" to-layer="1746" to-port="1" />
-		<edge from-layer="1505" from-port="0" to-layer="1506" to-port="1" />
-		<edge from-layer="1506" from-port="2" to-layer="1540" to-port="0" />
-		<edge from-layer="1506" from-port="2" to-layer="1509" to-port="0" />
-		<edge from-layer="1507" from-port="0" to-layer="1508" to-port="0" />
-		<edge from-layer="1508" from-port="1" to-layer="1509" to-port="1" />
-		<edge from-layer="1509" from-port="2" to-layer="1512" to-port="0" />
-		<edge from-layer="1510" from-port="0" to-layer="1511" to-port="0" />
-		<edge from-layer="1511" from-port="1" to-layer="1512" to-port="1" />
-		<edge from-layer="1512" from-port="2" to-layer="1521" to-port="0" />
-		<edge from-layer="1513" from-port="0" to-layer="1514" to-port="0" />
-		<edge from-layer="1514" from-port="1" to-layer="1517" to-port="0" />
-		<edge from-layer="1515" from-port="0" to-layer="1516" to-port="0" />
-		<edge from-layer="1516" from-port="1" to-layer="1517" to-port="1" />
-		<edge from-layer="1517" from-port="2" to-layer="1519" to-port="0" />
-		<edge from-layer="1518" from-port="0" to-layer="1519" to-port="1" />
-		<edge from-layer="1519" from-port="2" to-layer="1520" to-port="0" />
-		<edge from-layer="1520" from-port="1" to-layer="1521" to-port="1" />
-		<edge from-layer="1521" from-port="2" to-layer="1524" to-port="0" />
-		<edge from-layer="1522" from-port="0" to-layer="1523" to-port="0" />
-		<edge from-layer="1523" from-port="1" to-layer="1524" to-port="1" />
-		<edge from-layer="1524" from-port="2" to-layer="1525" to-port="0" />
-		<edge from-layer="1525" from-port="1" to-layer="1534" to-port="0" />
-		<edge from-layer="1526" from-port="0" to-layer="1527" to-port="0" />
-		<edge from-layer="1527" from-port="1" to-layer="1530" to-port="0" />
-		<edge from-layer="1528" from-port="0" to-layer="1529" to-port="0" />
-		<edge from-layer="1529" from-port="1" to-layer="1530" to-port="1" />
-		<edge from-layer="1530" from-port="2" to-layer="1532" to-port="0" />
-		<edge from-layer="1531" from-port="0" to-layer="1532" to-port="1" />
-		<edge from-layer="1532" from-port="2" to-layer="1533" to-port="0" />
-		<edge from-layer="1533" from-port="1" to-layer="1534" to-port="1" />
-		<edge from-layer="1534" from-port="2" to-layer="1537" to-port="0" />
-		<edge from-layer="1535" from-port="0" to-layer="1536" to-port="0" />
-		<edge from-layer="1536" from-port="1" to-layer="1537" to-port="1" />
-		<edge from-layer="1537" from-port="2" to-layer="1745" to-port="0" />
-		<edge from-layer="1538" from-port="0" to-layer="1539" to-port="0" />
-		<edge from-layer="1539" from-port="1" to-layer="1540" to-port="1" />
-		<edge from-layer="1540" from-port="2" to-layer="1543" to-port="0" />
-		<edge from-layer="1541" from-port="0" to-layer="1542" to-port="0" />
-		<edge from-layer="1542" from-port="1" to-layer="1543" to-port="1" />
-		<edge from-layer="1543" from-port="2" to-layer="1552" to-port="0" />
+		<edge from-layer="1488" from-port="2" to-layer="1492" to-port="0" />
+		<edge from-layer="1488" from-port="2" to-layer="1510" to-port="0" />
+		<edge from-layer="1489" from-port="0" to-layer="1492" to-port="1" />
+		<edge from-layer="1490" from-port="0" to-layer="1492" to-port="2" />
+		<edge from-layer="1491" from-port="0" to-layer="1492" to-port="3" />
+		<edge from-layer="1492" from-port="5" to-layer="1503" to-port="0" />
+		<edge from-layer="1492" from-port="5" to-layer="1493" to-port="0" />
+		<edge from-layer="1492" from-port="5" to-layer="1497" to-port="0" />
+		<edge from-layer="1493" from-port="2" to-layer="1506" to-port="0" />
+		<edge from-layer="1494" from-port="0" to-layer="1497" to-port="1" />
+		<edge from-layer="1495" from-port="0" to-layer="1497" to-port="2" />
+		<edge from-layer="1496" from-port="0" to-layer="1497" to-port="3" />
+		<edge from-layer="1497" from-port="5" to-layer="1499" to-port="0" />
+		<edge from-layer="1498" from-port="0" to-layer="1499" to-port="1" />
+		<edge from-layer="1499" from-port="2" to-layer="1504" to-port="0" />
+		<edge from-layer="1500" from-port="0" to-layer="1503" to-port="1" />
+		<edge from-layer="1501" from-port="0" to-layer="1503" to-port="2" />
+		<edge from-layer="1502" from-port="0" to-layer="1503" to-port="3" />
+		<edge from-layer="1503" from-port="5" to-layer="1504" to-port="1" />
+		<edge from-layer="1504" from-port="2" to-layer="1505" to-port="0" />
+		<edge from-layer="1505" from-port="2" to-layer="1506" to-port="1" />
+		<edge from-layer="1506" from-port="2" to-layer="1511" to-port="0" />
+		<edge from-layer="1507" from-port="0" to-layer="1510" to-port="1" />
+		<edge from-layer="1508" from-port="0" to-layer="1510" to-port="2" />
+		<edge from-layer="1509" from-port="0" to-layer="1510" to-port="3" />
+		<edge from-layer="1510" from-port="5" to-layer="1511" to-port="1" />
+		<edge from-layer="1511" from-port="2" to-layer="1512" to-port="1" />
+		<edge from-layer="1512" from-port="2" to-layer="1537" to-port="1" />
+		<edge from-layer="1512" from-port="2" to-layer="1623" to-port="0" />
+		<edge from-layer="1513" from-port="0" to-layer="1518" to-port="0" />
+		<edge from-layer="1514" from-port="0" to-layer="1517" to-port="1" />
+		<edge from-layer="1515" from-port="0" to-layer="1517" to-port="2" />
+		<edge from-layer="1516" from-port="0" to-layer="1517" to-port="3" />
+		<edge from-layer="1517" from-port="4" to-layer="1518" to-port="1" />
+		<edge from-layer="1518" from-port="2" to-layer="1519" to-port="0" />
+		<edge from-layer="1519" from-port="1" to-layer="1521" to-port="0" />
+		<edge from-layer="1520" from-port="0" to-layer="1521" to-port="2" />
+		<edge from-layer="1521" from-port="3" to-layer="1528" to-port="0" />
+		<edge from-layer="1522" from-port="0" to-layer="1525" to-port="1" />
+		<edge from-layer="1523" from-port="0" to-layer="1525" to-port="2" />
+		<edge from-layer="1524" from-port="0" to-layer="1525" to-port="3" />
+		<edge from-layer="1525" from-port="5" to-layer="1527" to-port="0" />
+		<edge from-layer="1526" from-port="0" to-layer="1527" to-port="1" />
+		<edge from-layer="1527" from-port="2" to-layer="1528" to-port="1" />
+		<edge from-layer="1528" from-port="2" to-layer="1622" to-port="0" />
+		<edge from-layer="1528" from-port="2" to-layer="1537" to-port="2" />
+		<edge from-layer="1529" from-port="0" to-layer="1536" to-port="1" />
+		<edge from-layer="1530" from-port="1" to-layer="1533" to-port="0" />
+		<edge from-layer="1531" from-port="0" to-layer="1533" to-port="1" />
+		<edge from-layer="1532" from-port="0" to-layer="1533" to-port="2" />
+		<edge from-layer="1533" from-port="3" to-layer="1534" to-port="0" />
+		<edge from-layer="1534" from-port="2" to-layer="1536" to-port="2" />
+		<edge from-layer="1535" from-port="0" to-layer="1536" to-port="3" />
+		<edge from-layer="1536" from-port="5" to-layer="1537" to-port="3" />
+		<edge from-layer="1537" from-port="4" to-layer="1539" to-port="0" />
+		<edge from-layer="1538" from-port="0" to-layer="1539" to-port="1" />
+		<edge from-layer="1539" from-port="2" to-layer="1541" to-port="0" />
+		<edge from-layer="1540" from-port="0" to-layer="1541" to-port="1" />
+		<edge from-layer="1541" from-port="2" to-layer="1550" to-port="0" />
+		<edge from-layer="1542" from-port="0" to-layer="1543" to-port="0" />
+		<edge from-layer="1543" from-port="1" to-layer="1546" to-port="0" />
 		<edge from-layer="1544" from-port="0" to-layer="1545" to-port="0" />
-		<edge from-layer="1545" from-port="1" to-layer="1548" to-port="0" />
-		<edge from-layer="1546" from-port="0" to-layer="1547" to-port="0" />
-		<edge from-layer="1547" from-port="1" to-layer="1548" to-port="1" />
-		<edge from-layer="1548" from-port="2" to-layer="1550" to-port="0" />
-		<edge from-layer="1549" from-port="0" to-layer="1550" to-port="1" />
-		<edge from-layer="1550" from-port="2" to-layer="1551" to-port="0" />
-		<edge from-layer="1551" from-port="1" to-layer="1552" to-port="1" />
-		<edge from-layer="1552" from-port="2" to-layer="1555" to-port="0" />
-		<edge from-layer="1553" from-port="0" to-layer="1554" to-port="0" />
-		<edge from-layer="1554" from-port="1" to-layer="1555" to-port="1" />
-		<edge from-layer="1555" from-port="2" to-layer="1557" to-port="0" />
-		<edge from-layer="1556" from-port="0" to-layer="1557" to-port="1" />
-		<edge from-layer="1557" from-port="2" to-layer="1561" to-port="0" />
-		<edge from-layer="1557" from-port="2" to-layer="1685" to-port="0" />
-		<edge from-layer="1557" from-port="2" to-layer="1574" to-port="0" />
-		<edge from-layer="1558" from-port="0" to-layer="1561" to-port="1" />
-		<edge from-layer="1559" from-port="0" to-layer="1561" to-port="2" />
-		<edge from-layer="1560" from-port="0" to-layer="1561" to-port="3" />
-		<edge from-layer="1561" from-port="4" to-layer="1563" to-port="0" />
-		<edge from-layer="1562" from-port="0" to-layer="1563" to-port="1" />
-		<edge from-layer="1563" from-port="2" to-layer="1567" to-port="0" />
-		<edge from-layer="1563" from-port="2" to-layer="1636" to-port="0" />
-		<edge from-layer="1564" from-port="0" to-layer="1567" to-port="1" />
-		<edge from-layer="1565" from-port="0" to-layer="1567" to-port="2" />
-		<edge from-layer="1566" from-port="0" to-layer="1567" to-port="3" />
-		<edge from-layer="1567" from-port="4" to-layer="1601" to-port="0" />
-		<edge from-layer="1567" from-port="4" to-layer="1604" to-port="0" />
-		<edge from-layer="1567" from-port="4" to-layer="1615" to-port="0" />
-		<edge from-layer="1567" from-port="4" to-layer="1621" to-port="0" />
-		<edge from-layer="1568" from-port="0" to-layer="1569" to-port="0" />
-		<edge from-layer="1569" from-port="1" to-layer="1597" to-port="0" />
-		<edge from-layer="1570" from-port="0" to-layer="1597" to-port="1" />
-		<edge from-layer="1571" from-port="0" to-layer="1574" to-port="1" />
-		<edge from-layer="1572" from-port="0" to-layer="1574" to-port="2" />
-		<edge from-layer="1573" from-port="0" to-layer="1574" to-port="3" />
-		<edge from-layer="1574" from-port="4" to-layer="1576" to-port="0" />
-		<edge from-layer="1575" from-port="0" to-layer="1576" to-port="1" />
-		<edge from-layer="1576" from-port="2" to-layer="1641" to-port="0" />
-		<edge from-layer="1576" from-port="2" to-layer="1669" to-port="0" />
-		<edge from-layer="1576" from-port="2" to-layer="1577" to-port="0" />
-		<edge from-layer="1577" from-port="1" to-layer="1579" to-port="0" />
-		<edge from-layer="1578" from-port="0" to-layer="1579" to-port="2" />
-		<edge from-layer="1579" from-port="3" to-layer="1593" to-port="0" />
-		<edge from-layer="1580" from-port="0" to-layer="1581" to-port="0" />
-		<edge from-layer="1581" from-port="1" to-layer="1586" to-port="0" />
-		<edge from-layer="1582" from-port="0" to-layer="1585" to-port="1" />
-		<edge from-layer="1583" from-port="0" to-layer="1585" to-port="2" />
-		<edge from-layer="1584" from-port="0" to-layer="1585" to-port="3" />
-		<edge from-layer="1585" from-port="4" to-layer="1586" to-port="1" />
-		<edge from-layer="1586" from-port="2" to-layer="1587" to-port="0" />
-		<edge from-layer="1587" from-port="1" to-layer="1589" to-port="0" />
-		<edge from-layer="1588" from-port="0" to-layer="1589" to-port="2" />
-		<edge from-layer="1589" from-port="3" to-layer="1671" to-port="0" />
-		<edge from-layer="1589" from-port="3" to-layer="1590" to-port="0" />
-		<edge from-layer="1590" from-port="1" to-layer="1592" to-port="0" />
-		<edge from-layer="1591" from-port="0" to-layer="1592" to-port="2" />
-		<edge from-layer="1592" from-port="3" to-layer="1593" to-port="1" />
-		<edge from-layer="1593" from-port="2" to-layer="1595" to-port="0" />
-		<edge from-layer="1594" from-port="0" to-layer="1595" to-port="1" />
-		<edge from-layer="1595" from-port="2" to-layer="1627" to-port="2" />
-		<edge from-layer="1595" from-port="2" to-layer="1597" to-port="2" />
-		<edge from-layer="1596" from-port="0" to-layer="1597" to-port="3" />
-		<edge from-layer="1597" from-port="4" to-layer="1599" to-port="0" />
-		<edge from-layer="1598" from-port="0" to-layer="1599" to-port="2" />
-		<edge from-layer="1599" from-port="3" to-layer="1600" to-port="0" />
-		<edge from-layer="1600" from-port="2" to-layer="1642" to-port="1" />
-		<edge from-layer="1600" from-port="2" to-layer="1601" to-port="1" />
-		<edge from-layer="1601" from-port="2" to-layer="1632" to-port="0" />
-		<edge from-layer="1602" from-port="0" to-layer="1612" to-port="0" />
-		<edge from-layer="1603" from-port="0" to-layer="1612" to-port="1" />
-		<edge from-layer="1604" from-port="1" to-layer="1606" to-port="0" />
-		<edge from-layer="1605" from-port="0" to-layer="1606" to-port="2" />
-		<edge from-layer="1606" from-port="3" to-layer="1607" to-port="0" />
-		<edge from-layer="1607" from-port="2" to-layer="1608" to-port="0" />
-		<edge from-layer="1608" from-port="1" to-layer="1610" to-port="0" />
-		<edge from-layer="1609" from-port="0" to-layer="1610" to-port="1" />
-		<edge from-layer="1610" from-port="2" to-layer="1612" to-port="2" />
-		<edge from-layer="1611" from-port="0" to-layer="1612" to-port="3" />
-		<edge from-layer="1612" from-port="4" to-layer="1615" to-port="1" />
-		<edge from-layer="1612" from-port="4" to-layer="1621" to-port="2" />
-		<edge from-layer="1613" from-port="0" to-layer="1615" to-port="2" />
-		<edge from-layer="1614" from-port="0" to-layer="1615" to-port="3" />
-		<edge from-layer="1615" from-port="4" to-layer="1618" to-port="0" />
-		<edge from-layer="1616" from-port="0" to-layer="1617" to-port="0" />
-		<edge from-layer="1617" from-port="1" to-layer="1618" to-port="1" />
-		<edge from-layer="1618" from-port="2" to-layer="1622" to-port="0" />
-		<edge from-layer="1619" from-port="0" to-layer="1621" to-port="1" />
-		<edge from-layer="1620" from-port="0" to-layer="1621" to-port="3" />
-		<edge from-layer="1621" from-port="4" to-layer="1622" to-port="1" />
-		<edge from-layer="1622" from-port="2" to-layer="1631" to-port="0" />
-		<edge from-layer="1623" from-port="0" to-layer="1624" to-port="0" />
-		<edge from-layer="1624" from-port="1" to-layer="1627" to-port="0" />
-		<edge from-layer="1625" from-port="0" to-layer="1627" to-port="1" />
-		<edge from-layer="1626" from-port="0" to-layer="1627" to-port="3" />
-		<edge from-layer="1627" from-port="4" to-layer="1629" to-port="0" />
-		<edge from-layer="1628" from-port="0" to-layer="1629" to-port="2" />
-		<edge from-layer="1629" from-port="3" to-layer="1630" to-port="0" />
-		<edge from-layer="1630" from-port="2" to-layer="1631" to-port="1" />
-		<edge from-layer="1630" from-port="2" to-layer="1664" to-port="1" />
-		<edge from-layer="1631" from-port="2" to-layer="1632" to-port="1" />
-		<edge from-layer="1632" from-port="2" to-layer="1637" to-port="0" />
+		<edge from-layer="1545" from-port="1" to-layer="1546" to-port="1" />
+		<edge from-layer="1546" from-port="2" to-layer="1548" to-port="0" />
+		<edge from-layer="1547" from-port="0" to-layer="1548" to-port="1" />
+		<edge from-layer="1548" from-port="2" to-layer="1549" to-port="0" />
+		<edge from-layer="1549" from-port="1" to-layer="1550" to-port="1" />
+		<edge from-layer="1550" from-port="2" to-layer="1552" to-port="0" />
+		<edge from-layer="1551" from-port="0" to-layer="1552" to-port="1" />
+		<edge from-layer="1552" from-port="2" to-layer="1553" to-port="1" />
+		<edge from-layer="1553" from-port="2" to-layer="1554" to-port="0" />
+		<edge from-layer="1554" from-port="2" to-layer="1724" to-port="1" />
+		<edge from-layer="1554" from-port="2" to-layer="1556" to-port="0" />
+		<edge from-layer="1555" from-port="0" to-layer="1556" to-port="1" />
+		<edge from-layer="1556" from-port="2" to-layer="1558" to-port="0" />
+		<edge from-layer="1556" from-port="2" to-layer="1642" to-port="0" />
+		<edge from-layer="1557" from-port="0" to-layer="1558" to-port="1" />
+		<edge from-layer="1558" from-port="2" to-layer="1560" to-port="0" />
+		<edge from-layer="1559" from-port="0" to-layer="1560" to-port="1" />
+		<edge from-layer="1560" from-port="2" to-layer="1569" to-port="0" />
+		<edge from-layer="1561" from-port="0" to-layer="1562" to-port="0" />
+		<edge from-layer="1562" from-port="1" to-layer="1565" to-port="0" />
+		<edge from-layer="1563" from-port="0" to-layer="1564" to-port="0" />
+		<edge from-layer="1564" from-port="1" to-layer="1565" to-port="1" />
+		<edge from-layer="1565" from-port="2" to-layer="1567" to-port="0" />
+		<edge from-layer="1566" from-port="0" to-layer="1567" to-port="1" />
+		<edge from-layer="1567" from-port="2" to-layer="1568" to-port="0" />
+		<edge from-layer="1568" from-port="1" to-layer="1569" to-port="1" />
+		<edge from-layer="1569" from-port="2" to-layer="1571" to-port="0" />
+		<edge from-layer="1570" from-port="0" to-layer="1571" to-port="1" />
+		<edge from-layer="1571" from-port="2" to-layer="1573" to-port="0" />
+		<edge from-layer="1572" from-port="0" to-layer="1573" to-port="1" />
+		<edge from-layer="1573" from-port="2" to-layer="1673" to-port="0" />
+		<edge from-layer="1573" from-port="2" to-layer="1594" to-port="0" />
+		<edge from-layer="1573" from-port="2" to-layer="1577" to-port="0" />
+		<edge from-layer="1574" from-port="0" to-layer="1577" to-port="1" />
+		<edge from-layer="1575" from-port="0" to-layer="1577" to-port="2" />
+		<edge from-layer="1576" from-port="0" to-layer="1577" to-port="3" />
+		<edge from-layer="1577" from-port="5" to-layer="1579" to-port="0" />
+		<edge from-layer="1578" from-port="0" to-layer="1579" to-port="1" />
+		<edge from-layer="1579" from-port="2" to-layer="1580" to-port="1" />
+		<edge from-layer="1580" from-port="2" to-layer="1581" to-port="0" />
+		<edge from-layer="1580" from-port="2" to-layer="1707" to-port="2" />
+		<edge from-layer="1582" from-port="0" to-layer="1587" to-port="0" />
+		<edge from-layer="1583" from-port="0" to-layer="1586" to-port="1" />
+		<edge from-layer="1584" from-port="0" to-layer="1586" to-port="2" />
+		<edge from-layer="1585" from-port="0" to-layer="1586" to-port="3" />
+		<edge from-layer="1586" from-port="4" to-layer="1587" to-port="1" />
+		<edge from-layer="1587" from-port="2" to-layer="1588" to-port="0" />
+		<edge from-layer="1588" from-port="1" to-layer="1590" to-port="0" />
+		<edge from-layer="1589" from-port="0" to-layer="1590" to-port="2" />
+		<edge from-layer="1590" from-port="3" to-layer="1620" to-port="0" />
+		<edge from-layer="1590" from-port="3" to-layer="1700" to-port="0" />
+		<edge from-layer="1591" from-port="0" to-layer="1594" to-port="1" />
+		<edge from-layer="1592" from-port="0" to-layer="1594" to-port="2" />
+		<edge from-layer="1593" from-port="0" to-layer="1594" to-port="3" />
+		<edge from-layer="1594" from-port="5" to-layer="1596" to-port="0" />
+		<edge from-layer="1595" from-port="0" to-layer="1596" to-port="1" />
+		<edge from-layer="1596" from-port="2" to-layer="1600" to-port="0" />
+		<edge from-layer="1596" from-port="2" to-layer="1618" to-port="0" />
+		<edge from-layer="1597" from-port="0" to-layer="1600" to-port="1" />
+		<edge from-layer="1598" from-port="0" to-layer="1600" to-port="2" />
+		<edge from-layer="1599" from-port="0" to-layer="1600" to-port="3" />
+		<edge from-layer="1600" from-port="5" to-layer="1611" to-port="0" />
+		<edge from-layer="1600" from-port="5" to-layer="1605" to-port="0" />
+		<edge from-layer="1600" from-port="5" to-layer="1601" to-port="0" />
+		<edge from-layer="1601" from-port="2" to-layer="1614" to-port="0" />
+		<edge from-layer="1602" from-port="0" to-layer="1605" to-port="1" />
+		<edge from-layer="1603" from-port="0" to-layer="1605" to-port="2" />
+		<edge from-layer="1604" from-port="0" to-layer="1605" to-port="3" />
+		<edge from-layer="1605" from-port="5" to-layer="1607" to-port="0" />
+		<edge from-layer="1606" from-port="0" to-layer="1607" to-port="1" />
+		<edge from-layer="1607" from-port="2" to-layer="1612" to-port="0" />
+		<edge from-layer="1608" from-port="0" to-layer="1611" to-port="1" />
+		<edge from-layer="1609" from-port="0" to-layer="1611" to-port="2" />
+		<edge from-layer="1610" from-port="0" to-layer="1611" to-port="3" />
+		<edge from-layer="1611" from-port="5" to-layer="1612" to-port="1" />
+		<edge from-layer="1612" from-port="2" to-layer="1613" to-port="0" />
+		<edge from-layer="1613" from-port="2" to-layer="1614" to-port="1" />
+		<edge from-layer="1614" from-port="2" to-layer="1619" to-port="0" />
+		<edge from-layer="1615" from-port="0" to-layer="1618" to-port="1" />
+		<edge from-layer="1616" from-port="0" to-layer="1618" to-port="2" />
+		<edge from-layer="1617" from-port="0" to-layer="1618" to-port="3" />
+		<edge from-layer="1618" from-port="5" to-layer="1619" to-port="1" />
+		<edge from-layer="1619" from-port="2" to-layer="1620" to-port="1" />
+		<edge from-layer="1620" from-port="2" to-layer="1621" to-port="0" />
+		<edge from-layer="1620" from-port="2" to-layer="1707" to-port="1" />
+		<edge from-layer="1632" from-port="0" to-layer="1637" to-port="0" />
 		<edge from-layer="1633" from-port="0" to-layer="1636" to-port="1" />
 		<edge from-layer="1634" from-port="0" to-layer="1636" to-port="2" />
 		<edge from-layer="1635" from-port="0" to-layer="1636" to-port="3" />
 		<edge from-layer="1636" from-port="4" to-layer="1637" to-port="1" />
-		<edge from-layer="1637" from-port="2" to-layer="1728" to-port="0" />
-		<edge from-layer="1637" from-port="2" to-layer="1695" to-port="0" />
-		<edge from-layer="1638" from-port="0" to-layer="1641" to-port="1" />
-		<edge from-layer="1639" from-port="0" to-layer="1641" to-port="2" />
-		<edge from-layer="1640" from-port="0" to-layer="1641" to-port="3" />
-		<edge from-layer="1641" from-port="4" to-layer="1642" to-port="0" />
-		<edge from-layer="1641" from-port="4" to-layer="1645" to-port="0" />
-		<edge from-layer="1641" from-port="4" to-layer="1662" to-port="0" />
-		<edge from-layer="1641" from-port="4" to-layer="1656" to-port="0" />
-		<edge from-layer="1642" from-port="2" to-layer="1665" to-port="0" />
-		<edge from-layer="1643" from-port="0" to-layer="1653" to-port="0" />
-		<edge from-layer="1644" from-port="0" to-layer="1653" to-port="1" />
-		<edge from-layer="1645" from-port="1" to-layer="1647" to-port="0" />
-		<edge from-layer="1646" from-port="0" to-layer="1647" to-port="2" />
-		<edge from-layer="1647" from-port="3" to-layer="1648" to-port="0" />
-		<edge from-layer="1648" from-port="2" to-layer="1649" to-port="0" />
-		<edge from-layer="1649" from-port="1" to-layer="1651" to-port="0" />
+		<edge from-layer="1637" from-port="2" to-layer="1638" to-port="0" />
+		<edge from-layer="1638" from-port="1" to-layer="1640" to-port="0" />
+		<edge from-layer="1639" from-port="0" to-layer="1640" to-port="2" />
+		<edge from-layer="1640" from-port="3" to-layer="5068" to-port="0" />
+		<edge from-layer="1641" from-port="0" to-layer="1642" to-port="1" />
+		<edge from-layer="1642" from-port="2" to-layer="1644" to-port="0" />
+		<edge from-layer="1643" from-port="0" to-layer="1644" to-port="1" />
+		<edge from-layer="1644" from-port="2" to-layer="1653" to-port="0" />
+		<edge from-layer="1645" from-port="0" to-layer="1646" to-port="0" />
+		<edge from-layer="1646" from-port="1" to-layer="1649" to-port="0" />
+		<edge from-layer="1647" from-port="0" to-layer="1648" to-port="0" />
+		<edge from-layer="1648" from-port="1" to-layer="1649" to-port="1" />
+		<edge from-layer="1649" from-port="2" to-layer="1651" to-port="0" />
 		<edge from-layer="1650" from-port="0" to-layer="1651" to-port="1" />
-		<edge from-layer="1651" from-port="2" to-layer="1653" to-port="2" />
-		<edge from-layer="1652" from-port="0" to-layer="1653" to-port="3" />
-		<edge from-layer="1653" from-port="4" to-layer="1662" to-port="2" />
-		<edge from-layer="1653" from-port="4" to-layer="1656" to-port="1" />
-		<edge from-layer="1654" from-port="0" to-layer="1656" to-port="2" />
-		<edge from-layer="1655" from-port="0" to-layer="1656" to-port="3" />
-		<edge from-layer="1656" from-port="4" to-layer="1659" to-port="0" />
-		<edge from-layer="1657" from-port="0" to-layer="1658" to-port="0" />
-		<edge from-layer="1658" from-port="1" to-layer="1659" to-port="1" />
-		<edge from-layer="1659" from-port="2" to-layer="1663" to-port="0" />
-		<edge from-layer="1660" from-port="0" to-layer="1662" to-port="1" />
-		<edge from-layer="1661" from-port="0" to-layer="1662" to-port="3" />
-		<edge from-layer="1662" from-port="4" to-layer="1663" to-port="1" />
-		<edge from-layer="1663" from-port="2" to-layer="1664" to-port="0" />
-		<edge from-layer="1664" from-port="2" to-layer="1665" to-port="1" />
-		<edge from-layer="1665" from-port="2" to-layer="1670" to-port="0" />
-		<edge from-layer="1666" from-port="0" to-layer="1669" to-port="1" />
-		<edge from-layer="1667" from-port="0" to-layer="1669" to-port="2" />
-		<edge from-layer="1668" from-port="0" to-layer="1669" to-port="3" />
-		<edge from-layer="1669" from-port="4" to-layer="1670" to-port="1" />
-		<edge from-layer="1670" from-port="2" to-layer="1671" to-port="1" />
-		<edge from-layer="1671" from-port="2" to-layer="2342" to-port="0" />
-		<edge from-layer="1671" from-port="2" to-layer="1692" to-port="0" />
-		<edge from-layer="1671" from-port="2" to-layer="1728" to-port="1" />
-		<edge from-layer="1672" from-port="0" to-layer="1673" to-port="0" />
-		<edge from-layer="1673" from-port="1" to-layer="1678" to-port="0" />
-		<edge from-layer="1674" from-port="0" to-layer="1677" to-port="1" />
-		<edge from-layer="1675" from-port="0" to-layer="1677" to-port="2" />
-		<edge from-layer="1676" from-port="0" to-layer="1677" to-port="3" />
-		<edge from-layer="1677" from-port="4" to-layer="1678" to-port="1" />
-		<edge from-layer="1678" from-port="2" to-layer="1679" to-port="0" />
-		<edge from-layer="1679" from-port="1" to-layer="1681" to-port="0" />
-		<edge from-layer="1680" from-port="0" to-layer="1681" to-port="2" />
-		<edge from-layer="1681" from-port="3" to-layer="1688" to-port="0" />
-		<edge from-layer="1682" from-port="0" to-layer="1685" to-port="1" />
-		<edge from-layer="1683" from-port="0" to-layer="1685" to-port="2" />
-		<edge from-layer="1684" from-port="0" to-layer="1685" to-port="3" />
-		<edge from-layer="1685" from-port="4" to-layer="1687" to-port="0" />
-		<edge from-layer="1686" from-port="0" to-layer="1687" to-port="1" />
-		<edge from-layer="1687" from-port="2" to-layer="1688" to-port="1" />
-		<edge from-layer="1688" from-port="2" to-layer="1728" to-port="2" />
-		<edge from-layer="1688" from-port="2" to-layer="2341" to-port="0" />
-		<edge from-layer="1689" from-port="0" to-layer="1707" to-port="0" />
-		<edge from-layer="1690" from-port="0" to-layer="1702" to-port="0" />
-		<edge from-layer="1691" from-port="0" to-layer="1702" to-port="1" />
-		<edge from-layer="1691" from-port="0" to-layer="1705" to-port="1" />
-		<edge from-layer="1692" from-port="1" to-layer="1694" to-port="0" />
-		<edge from-layer="1693" from-port="0" to-layer="1694" to-port="2" />
-		<edge from-layer="1694" from-port="3" to-layer="1704" to-port="0" />
-		<edge from-layer="1694" from-port="3" to-layer="1698" to-port="0" />
-		<edge from-layer="1695" from-port="1" to-layer="1723" to-port="0" />
-		<edge from-layer="1695" from-port="1" to-layer="1697" to-port="0" />
-		<edge from-layer="1696" from-port="0" to-layer="1697" to-port="2" />
-		<edge from-layer="1697" from-port="3" to-layer="1698" to-port="1" />
-		<edge from-layer="1698" from-port="2" to-layer="1700" to-port="0" />
-		<edge from-layer="1699" from-port="0" to-layer="1704" to-port="1" />
-		<edge from-layer="1699" from-port="0" to-layer="1700" to-port="1" />
-		<edge from-layer="1700" from-port="2" to-layer="1702" to-port="2" />
-		<edge from-layer="1701" from-port="0" to-layer="1705" to-port="3" />
-		<edge from-layer="1701" from-port="0" to-layer="1702" to-port="3" />
-		<edge from-layer="1702" from-port="4" to-layer="1707" to-port="1" />
-		<edge from-layer="1703" from-port="0" to-layer="1705" to-port="0" />
-		<edge from-layer="1704" from-port="2" to-layer="1712" to-port="2" />
-		<edge from-layer="1704" from-port="2" to-layer="1705" to-port="2" />
-		<edge from-layer="1705" from-port="4" to-layer="1707" to-port="2" />
-		<edge from-layer="1706" from-port="0" to-layer="1707" to-port="3" />
-		<edge from-layer="1707" from-port="4" to-layer="1714" to-port="0" />
-		<edge from-layer="1708" from-port="0" to-layer="1714" to-port="1" />
-		<edge from-layer="1709" from-port="0" to-layer="1712" to-port="0" />
-		<edge from-layer="1710" from-port="0" to-layer="1712" to-port="1" />
-		<edge from-layer="1711" from-port="0" to-layer="1712" to-port="3" />
-		<edge from-layer="1712" from-port="4" to-layer="1714" to-port="2" />
-		<edge from-layer="1713" from-port="0" to-layer="1714" to-port="3" />
-		<edge from-layer="1714" from-port="4" to-layer="1718" to-port="0" />
-		<edge from-layer="1715" from-port="0" to-layer="1716" to-port="0" />
-		<edge from-layer="1716" from-port="1" to-layer="1718" to-port="1" />
-		<edge from-layer="1717" from-port="0" to-layer="1718" to-port="2" />
-		<edge from-layer="1718" from-port="3" to-layer="1719" to-port="0" />
-		<edge from-layer="1719" from-port="2" to-layer="1720" to-port="0" />
-		<edge from-layer="1719" from-port="2" to-layer="1727" to-port="0" />
-		<edge from-layer="1720" from-port="1" to-layer="1726" to-port="0" />
-		<edge from-layer="1721" from-port="0" to-layer="1723" to-port="1" />
-		<edge from-layer="1722" from-port="0" to-layer="1723" to-port="2" />
-		<edge from-layer="1723" from-port="3" to-layer="1724" to-port="0" />
-		<edge from-layer="1724" from-port="4" to-layer="1725" to-port="0" />
-		<edge from-layer="1725" from-port="1" to-layer="1726" to-port="1" />
-		<edge from-layer="1726" from-port="2" to-layer="1727" to-port="1" />
-		<edge from-layer="1727" from-port="2" to-layer="1728" to-port="3" />
-		<edge from-layer="1728" from-port="4" to-layer="1730" to-port="0" />
+		<edge from-layer="1651" from-port="2" to-layer="1652" to-port="0" />
+		<edge from-layer="1652" from-port="1" to-layer="1653" to-port="1" />
+		<edge from-layer="1653" from-port="2" to-layer="1656" to-port="0" />
+		<edge from-layer="1654" from-port="0" to-layer="1655" to-port="0" />
+		<edge from-layer="1655" from-port="1" to-layer="1656" to-port="1" />
+		<edge from-layer="1656" from-port="2" to-layer="1657" to-port="0" />
+		<edge from-layer="1657" from-port="1" to-layer="1666" to-port="0" />
+		<edge from-layer="1658" from-port="0" to-layer="1659" to-port="0" />
+		<edge from-layer="1659" from-port="1" to-layer="1662" to-port="0" />
+		<edge from-layer="1660" from-port="0" to-layer="1661" to-port="0" />
+		<edge from-layer="1661" from-port="1" to-layer="1662" to-port="1" />
+		<edge from-layer="1662" from-port="2" to-layer="1664" to-port="0" />
+		<edge from-layer="1663" from-port="0" to-layer="1664" to-port="1" />
+		<edge from-layer="1664" from-port="2" to-layer="1665" to-port="0" />
+		<edge from-layer="1665" from-port="1" to-layer="1666" to-port="1" />
+		<edge from-layer="1666" from-port="2" to-layer="1669" to-port="0" />
+		<edge from-layer="1667" from-port="0" to-layer="1668" to-port="0" />
+		<edge from-layer="1668" from-port="1" to-layer="1669" to-port="1" />
+		<edge from-layer="1669" from-port="2" to-layer="1723" to-port="0" />
+		<edge from-layer="1670" from-port="0" to-layer="1673" to-port="1" />
+		<edge from-layer="1671" from-port="0" to-layer="1673" to-port="2" />
+		<edge from-layer="1672" from-port="0" to-layer="1673" to-port="3" />
+		<edge from-layer="1673" from-port="5" to-layer="1675" to-port="0" />
+		<edge from-layer="1674" from-port="0" to-layer="1675" to-port="1" />
+		<edge from-layer="1675" from-port="2" to-layer="1697" to-port="0" />
+		<edge from-layer="1675" from-port="2" to-layer="1679" to-port="0" />
+		<edge from-layer="1676" from-port="0" to-layer="1679" to-port="1" />
+		<edge from-layer="1677" from-port="0" to-layer="1679" to-port="2" />
+		<edge from-layer="1678" from-port="0" to-layer="1679" to-port="3" />
+		<edge from-layer="1679" from-port="5" to-layer="1684" to-port="0" />
+		<edge from-layer="1679" from-port="5" to-layer="1680" to-port="0" />
+		<edge from-layer="1679" from-port="5" to-layer="1690" to-port="0" />
+		<edge from-layer="1680" from-port="2" to-layer="1693" to-port="0" />
+		<edge from-layer="1681" from-port="0" to-layer="1684" to-port="1" />
+		<edge from-layer="1682" from-port="0" to-layer="1684" to-port="2" />
+		<edge from-layer="1683" from-port="0" to-layer="1684" to-port="3" />
+		<edge from-layer="1684" from-port="5" to-layer="1686" to-port="0" />
+		<edge from-layer="1685" from-port="0" to-layer="1686" to-port="1" />
+		<edge from-layer="1686" from-port="2" to-layer="1691" to-port="0" />
+		<edge from-layer="1687" from-port="0" to-layer="1690" to-port="1" />
+		<edge from-layer="1688" from-port="0" to-layer="1690" to-port="2" />
+		<edge from-layer="1689" from-port="0" to-layer="1690" to-port="3" />
+		<edge from-layer="1690" from-port="5" to-layer="1691" to-port="1" />
+		<edge from-layer="1691" from-port="2" to-layer="1692" to-port="0" />
+		<edge from-layer="1692" from-port="2" to-layer="1693" to-port="1" />
+		<edge from-layer="1693" from-port="2" to-layer="1698" to-port="0" />
+		<edge from-layer="1694" from-port="0" to-layer="1697" to-port="1" />
+		<edge from-layer="1695" from-port="0" to-layer="1697" to-port="2" />
+		<edge from-layer="1696" from-port="0" to-layer="1697" to-port="3" />
+		<edge from-layer="1697" from-port="5" to-layer="1698" to-port="1" />
+		<edge from-layer="1698" from-port="2" to-layer="1707" to-port="0" />
+		<edge from-layer="1699" from-port="0" to-layer="1706" to-port="1" />
+		<edge from-layer="1700" from-port="1" to-layer="1703" to-port="0" />
+		<edge from-layer="1701" from-port="0" to-layer="1703" to-port="1" />
+		<edge from-layer="1702" from-port="0" to-layer="1703" to-port="2" />
+		<edge from-layer="1703" from-port="3" to-layer="1704" to-port="0" />
+		<edge from-layer="1704" from-port="2" to-layer="1706" to-port="2" />
+		<edge from-layer="1705" from-port="0" to-layer="1706" to-port="3" />
+		<edge from-layer="1706" from-port="5" to-layer="1707" to-port="3" />
+		<edge from-layer="1707" from-port="4" to-layer="1709" to-port="0" />
+		<edge from-layer="1708" from-port="0" to-layer="1709" to-port="1" />
+		<edge from-layer="1709" from-port="2" to-layer="1711" to-port="0" />
+		<edge from-layer="1710" from-port="0" to-layer="1711" to-port="1" />
+		<edge from-layer="1711" from-port="2" to-layer="1720" to-port="0" />
+		<edge from-layer="1712" from-port="0" to-layer="1713" to-port="0" />
+		<edge from-layer="1713" from-port="1" to-layer="1716" to-port="0" />
+		<edge from-layer="1714" from-port="0" to-layer="1715" to-port="0" />
+		<edge from-layer="1715" from-port="1" to-layer="1716" to-port="1" />
+		<edge from-layer="1716" from-port="2" to-layer="1718" to-port="0" />
+		<edge from-layer="1717" from-port="0" to-layer="1718" to-port="1" />
+		<edge from-layer="1718" from-port="2" to-layer="1719" to-port="0" />
+		<edge from-layer="1719" from-port="1" to-layer="1720" to-port="1" />
+		<edge from-layer="1720" from-port="2" to-layer="1722" to-port="0" />
+		<edge from-layer="1721" from-port="0" to-layer="1722" to-port="1" />
+		<edge from-layer="1722" from-port="2" to-layer="1723" to-port="1" />
+		<edge from-layer="1723" from-port="2" to-layer="1724" to-port="0" />
+		<edge from-layer="1724" from-port="2" to-layer="1726" to-port="0" />
+		<edge from-layer="1724" from-port="2" to-layer="1882" to-port="1" />
+		<edge from-layer="1725" from-port="0" to-layer="1726" to-port="1" />
+		<edge from-layer="1726" from-port="2" to-layer="1728" to-port="0" />
+		<edge from-layer="1726" from-port="2" to-layer="1757" to-port="0" />
+		<edge from-layer="1727" from-port="0" to-layer="1728" to-port="1" />
+		<edge from-layer="1728" from-port="2" to-layer="1730" to-port="0" />
 		<edge from-layer="1729" from-port="0" to-layer="1730" to-port="1" />
-		<edge from-layer="1730" from-port="2" to-layer="1732" to-port="0" />
-		<edge from-layer="1731" from-port="0" to-layer="1732" to-port="1" />
-		<edge from-layer="1732" from-port="2" to-layer="1741" to-port="0" />
+		<edge from-layer="1730" from-port="2" to-layer="1739" to-port="0" />
+		<edge from-layer="1731" from-port="0" to-layer="1732" to-port="0" />
+		<edge from-layer="1732" from-port="1" to-layer="1735" to-port="0" />
 		<edge from-layer="1733" from-port="0" to-layer="1734" to-port="0" />
-		<edge from-layer="1734" from-port="1" to-layer="1737" to-port="0" />
-		<edge from-layer="1735" from-port="0" to-layer="1736" to-port="0" />
-		<edge from-layer="1736" from-port="1" to-layer="1737" to-port="1" />
-		<edge from-layer="1737" from-port="2" to-layer="1739" to-port="0" />
-		<edge from-layer="1738" from-port="0" to-layer="1739" to-port="1" />
-		<edge from-layer="1739" from-port="2" to-layer="1740" to-port="0" />
-		<edge from-layer="1740" from-port="1" to-layer="1741" to-port="1" />
-		<edge from-layer="1741" from-port="2" to-layer="1744" to-port="0" />
-		<edge from-layer="1742" from-port="0" to-layer="1743" to-port="0" />
-		<edge from-layer="1743" from-port="1" to-layer="1744" to-port="1" />
-		<edge from-layer="1744" from-port="2" to-layer="1745" to-port="1" />
-		<edge from-layer="1745" from-port="2" to-layer="1746" to-port="0" />
-		<edge from-layer="1746" from-port="2" to-layer="1748" to-port="0" />
-		<edge from-layer="1746" from-port="2" to-layer="1988" to-port="1" />
-		<edge from-layer="1747" from-port="0" to-layer="1748" to-port="1" />
-		<edge from-layer="1748" from-port="2" to-layer="1782" to-port="0" />
-		<edge from-layer="1748" from-port="2" to-layer="1751" to-port="0" />
-		<edge from-layer="1749" from-port="0" to-layer="1750" to-port="0" />
-		<edge from-layer="1750" from-port="1" to-layer="1751" to-port="1" />
-		<edge from-layer="1751" from-port="2" to-layer="1754" to-port="0" />
-		<edge from-layer="1752" from-port="0" to-layer="1753" to-port="0" />
-		<edge from-layer="1753" from-port="1" to-layer="1754" to-port="1" />
-		<edge from-layer="1754" from-port="2" to-layer="1763" to-port="0" />
-		<edge from-layer="1755" from-port="0" to-layer="1756" to-port="0" />
-		<edge from-layer="1756" from-port="1" to-layer="1759" to-port="0" />
-		<edge from-layer="1757" from-port="0" to-layer="1758" to-port="0" />
-		<edge from-layer="1758" from-port="1" to-layer="1759" to-port="1" />
-		<edge from-layer="1759" from-port="2" to-layer="1761" to-port="0" />
-		<edge from-layer="1760" from-port="0" to-layer="1761" to-port="1" />
-		<edge from-layer="1761" from-port="2" to-layer="1762" to-port="0" />
-		<edge from-layer="1762" from-port="1" to-layer="1763" to-port="1" />
-		<edge from-layer="1763" from-port="2" to-layer="1766" to-port="0" />
-		<edge from-layer="1764" from-port="0" to-layer="1765" to-port="0" />
-		<edge from-layer="1765" from-port="1" to-layer="1766" to-port="1" />
+		<edge from-layer="1734" from-port="1" to-layer="1735" to-port="1" />
+		<edge from-layer="1735" from-port="2" to-layer="1737" to-port="0" />
+		<edge from-layer="1736" from-port="0" to-layer="1737" to-port="1" />
+		<edge from-layer="1737" from-port="2" to-layer="1738" to-port="0" />
+		<edge from-layer="1738" from-port="1" to-layer="1739" to-port="1" />
+		<edge from-layer="1739" from-port="2" to-layer="1742" to-port="0" />
+		<edge from-layer="1740" from-port="0" to-layer="1741" to-port="0" />
+		<edge from-layer="1741" from-port="1" to-layer="1742" to-port="1" />
+		<edge from-layer="1742" from-port="2" to-layer="1743" to-port="0" />
+		<edge from-layer="1743" from-port="1" to-layer="1752" to-port="0" />
+		<edge from-layer="1744" from-port="0" to-layer="1745" to-port="0" />
+		<edge from-layer="1745" from-port="1" to-layer="1748" to-port="0" />
+		<edge from-layer="1746" from-port="0" to-layer="1747" to-port="0" />
+		<edge from-layer="1747" from-port="1" to-layer="1748" to-port="1" />
+		<edge from-layer="1748" from-port="2" to-layer="1750" to-port="0" />
+		<edge from-layer="1749" from-port="0" to-layer="1750" to-port="1" />
+		<edge from-layer="1750" from-port="2" to-layer="1751" to-port="0" />
+		<edge from-layer="1751" from-port="1" to-layer="1752" to-port="1" />
+		<edge from-layer="1752" from-port="2" to-layer="1755" to-port="0" />
+		<edge from-layer="1753" from-port="0" to-layer="1754" to-port="0" />
+		<edge from-layer="1754" from-port="1" to-layer="1755" to-port="1" />
+		<edge from-layer="1755" from-port="2" to-layer="1881" to-port="0" />
+		<edge from-layer="1756" from-port="0" to-layer="1757" to-port="1" />
+		<edge from-layer="1757" from-port="2" to-layer="1759" to-port="0" />
+		<edge from-layer="1758" from-port="0" to-layer="1759" to-port="1" />
+		<edge from-layer="1759" from-port="2" to-layer="1768" to-port="0" />
+		<edge from-layer="1760" from-port="0" to-layer="1761" to-port="0" />
+		<edge from-layer="1761" from-port="1" to-layer="1764" to-port="0" />
+		<edge from-layer="1762" from-port="0" to-layer="1763" to-port="0" />
+		<edge from-layer="1763" from-port="1" to-layer="1764" to-port="1" />
+		<edge from-layer="1764" from-port="2" to-layer="1766" to-port="0" />
+		<edge from-layer="1765" from-port="0" to-layer="1766" to-port="1" />
 		<edge from-layer="1766" from-port="2" to-layer="1767" to-port="0" />
-		<edge from-layer="1767" from-port="1" to-layer="1776" to-port="0" />
-		<edge from-layer="1768" from-port="0" to-layer="1769" to-port="0" />
-		<edge from-layer="1769" from-port="1" to-layer="1772" to-port="0" />
-		<edge from-layer="1770" from-port="0" to-layer="1771" to-port="0" />
-		<edge from-layer="1771" from-port="1" to-layer="1772" to-port="1" />
-		<edge from-layer="1772" from-port="2" to-layer="1774" to-port="0" />
-		<edge from-layer="1773" from-port="0" to-layer="1774" to-port="1" />
-		<edge from-layer="1774" from-port="2" to-layer="1775" to-port="0" />
-		<edge from-layer="1775" from-port="1" to-layer="1776" to-port="1" />
-		<edge from-layer="1776" from-port="2" to-layer="1779" to-port="0" />
-		<edge from-layer="1777" from-port="0" to-layer="1778" to-port="0" />
-		<edge from-layer="1778" from-port="1" to-layer="1779" to-port="1" />
-		<edge from-layer="1779" from-port="2" to-layer="1987" to-port="0" />
-		<edge from-layer="1780" from-port="0" to-layer="1781" to-port="0" />
-		<edge from-layer="1781" from-port="1" to-layer="1782" to-port="1" />
-		<edge from-layer="1782" from-port="2" to-layer="1785" to-port="0" />
-		<edge from-layer="1783" from-port="0" to-layer="1784" to-port="0" />
-		<edge from-layer="1784" from-port="1" to-layer="1785" to-port="1" />
-		<edge from-layer="1785" from-port="2" to-layer="1794" to-port="0" />
-		<edge from-layer="1786" from-port="0" to-layer="1787" to-port="0" />
-		<edge from-layer="1787" from-port="1" to-layer="1790" to-port="0" />
-		<edge from-layer="1788" from-port="0" to-layer="1789" to-port="0" />
-		<edge from-layer="1789" from-port="1" to-layer="1790" to-port="1" />
-		<edge from-layer="1790" from-port="2" to-layer="1792" to-port="0" />
-		<edge from-layer="1791" from-port="0" to-layer="1792" to-port="1" />
-		<edge from-layer="1792" from-port="2" to-layer="1793" to-port="0" />
-		<edge from-layer="1793" from-port="1" to-layer="1794" to-port="1" />
-		<edge from-layer="1794" from-port="2" to-layer="1797" to-port="0" />
-		<edge from-layer="1795" from-port="0" to-layer="1796" to-port="0" />
-		<edge from-layer="1796" from-port="1" to-layer="1797" to-port="1" />
-		<edge from-layer="1797" from-port="2" to-layer="1799" to-port="0" />
-		<edge from-layer="1798" from-port="0" to-layer="1799" to-port="1" />
-		<edge from-layer="1799" from-port="2" to-layer="1803" to-port="0" />
-		<edge from-layer="1799" from-port="2" to-layer="1927" to-port="0" />
-		<edge from-layer="1799" from-port="2" to-layer="1816" to-port="0" />
-		<edge from-layer="1800" from-port="0" to-layer="1803" to-port="1" />
-		<edge from-layer="1801" from-port="0" to-layer="1803" to-port="2" />
-		<edge from-layer="1802" from-port="0" to-layer="1803" to-port="3" />
-		<edge from-layer="1803" from-port="4" to-layer="1805" to-port="0" />
-		<edge from-layer="1804" from-port="0" to-layer="1805" to-port="1" />
-		<edge from-layer="1805" from-port="2" to-layer="1878" to-port="0" />
-		<edge from-layer="1805" from-port="2" to-layer="1809" to-port="0" />
-		<edge from-layer="1806" from-port="0" to-layer="1809" to-port="1" />
-		<edge from-layer="1807" from-port="0" to-layer="1809" to-port="2" />
-		<edge from-layer="1808" from-port="0" to-layer="1809" to-port="3" />
-		<edge from-layer="1809" from-port="4" to-layer="1857" to-port="0" />
-		<edge from-layer="1809" from-port="4" to-layer="1863" to-port="0" />
-		<edge from-layer="1809" from-port="4" to-layer="1846" to-port="0" />
-		<edge from-layer="1809" from-port="4" to-layer="1843" to-port="0" />
-		<edge from-layer="1810" from-port="0" to-layer="1811" to-port="0" />
-		<edge from-layer="1811" from-port="1" to-layer="1839" to-port="0" />
-		<edge from-layer="1812" from-port="0" to-layer="1839" to-port="1" />
-		<edge from-layer="1813" from-port="0" to-layer="1816" to-port="1" />
-		<edge from-layer="1814" from-port="0" to-layer="1816" to-port="2" />
-		<edge from-layer="1815" from-port="0" to-layer="1816" to-port="3" />
-		<edge from-layer="1816" from-port="4" to-layer="1818" to-port="0" />
-		<edge from-layer="1817" from-port="0" to-layer="1818" to-port="1" />
-		<edge from-layer="1818" from-port="2" to-layer="1883" to-port="0" />
-		<edge from-layer="1818" from-port="2" to-layer="1819" to-port="0" />
-		<edge from-layer="1818" from-port="2" to-layer="1911" to-port="0" />
-		<edge from-layer="1819" from-port="1" to-layer="1821" to-port="0" />
-		<edge from-layer="1820" from-port="0" to-layer="1821" to-port="2" />
-		<edge from-layer="1821" from-port="3" to-layer="1835" to-port="0" />
-		<edge from-layer="1822" from-port="0" to-layer="1823" to-port="0" />
-		<edge from-layer="1823" from-port="1" to-layer="1828" to-port="0" />
-		<edge from-layer="1824" from-port="0" to-layer="1827" to-port="1" />
-		<edge from-layer="1825" from-port="0" to-layer="1827" to-port="2" />
-		<edge from-layer="1826" from-port="0" to-layer="1827" to-port="3" />
-		<edge from-layer="1827" from-port="4" to-layer="1828" to-port="1" />
-		<edge from-layer="1828" from-port="2" to-layer="1829" to-port="0" />
-		<edge from-layer="1829" from-port="1" to-layer="1831" to-port="0" />
-		<edge from-layer="1830" from-port="0" to-layer="1831" to-port="2" />
-		<edge from-layer="1831" from-port="3" to-layer="1913" to-port="0" />
-		<edge from-layer="1831" from-port="3" to-layer="1832" to-port="0" />
-		<edge from-layer="1832" from-port="1" to-layer="1834" to-port="0" />
-		<edge from-layer="1833" from-port="0" to-layer="1834" to-port="2" />
-		<edge from-layer="1834" from-port="3" to-layer="1835" to-port="1" />
-		<edge from-layer="1835" from-port="2" to-layer="1837" to-port="0" />
-		<edge from-layer="1836" from-port="0" to-layer="1837" to-port="1" />
-		<edge from-layer="1837" from-port="2" to-layer="1869" to-port="2" />
-		<edge from-layer="1837" from-port="2" to-layer="1839" to-port="2" />
-		<edge from-layer="1838" from-port="0" to-layer="1839" to-port="3" />
-		<edge from-layer="1839" from-port="4" to-layer="1841" to-port="0" />
-		<edge from-layer="1840" from-port="0" to-layer="1841" to-port="2" />
-		<edge from-layer="1841" from-port="3" to-layer="1842" to-port="0" />
-		<edge from-layer="1842" from-port="2" to-layer="1884" to-port="1" />
-		<edge from-layer="1842" from-port="2" to-layer="1843" to-port="1" />
-		<edge from-layer="1843" from-port="2" to-layer="1874" to-port="0" />
-		<edge from-layer="1844" from-port="0" to-layer="1854" to-port="0" />
-		<edge from-layer="1845" from-port="0" to-layer="1854" to-port="1" />
-		<edge from-layer="1846" from-port="1" to-layer="1848" to-port="0" />
-		<edge from-layer="1847" from-port="0" to-layer="1848" to-port="2" />
-		<edge from-layer="1848" from-port="3" to-layer="1849" to-port="0" />
-		<edge from-layer="1849" from-port="2" to-layer="1850" to-port="0" />
-		<edge from-layer="1850" from-port="1" to-layer="1852" to-port="0" />
-		<edge from-layer="1851" from-port="0" to-layer="1852" to-port="1" />
-		<edge from-layer="1852" from-port="2" to-layer="1854" to-port="2" />
-		<edge from-layer="1853" from-port="0" to-layer="1854" to-port="3" />
-		<edge from-layer="1854" from-port="4" to-layer="1863" to-port="2" />
-		<edge from-layer="1854" from-port="4" to-layer="1857" to-port="1" />
-		<edge from-layer="1855" from-port="0" to-layer="1857" to-port="2" />
-		<edge from-layer="1856" from-port="0" to-layer="1857" to-port="3" />
-		<edge from-layer="1857" from-port="4" to-layer="1860" to-port="0" />
-		<edge from-layer="1858" from-port="0" to-layer="1859" to-port="0" />
-		<edge from-layer="1859" from-port="1" to-layer="1860" to-port="1" />
-		<edge from-layer="1860" from-port="2" to-layer="1864" to-port="0" />
-		<edge from-layer="1861" from-port="0" to-layer="1863" to-port="1" />
-		<edge from-layer="1862" from-port="0" to-layer="1863" to-port="3" />
-		<edge from-layer="1863" from-port="4" to-layer="1864" to-port="1" />
-		<edge from-layer="1864" from-port="2" to-layer="1873" to-port="0" />
-		<edge from-layer="1865" from-port="0" to-layer="1866" to-port="0" />
-		<edge from-layer="1866" from-port="1" to-layer="1869" to-port="0" />
-		<edge from-layer="1867" from-port="0" to-layer="1869" to-port="1" />
-		<edge from-layer="1868" from-port="0" to-layer="1869" to-port="3" />
-		<edge from-layer="1869" from-port="4" to-layer="1871" to-port="0" />
-		<edge from-layer="1870" from-port="0" to-layer="1871" to-port="2" />
-		<edge from-layer="1871" from-port="3" to-layer="1872" to-port="0" />
-		<edge from-layer="1872" from-port="2" to-layer="1873" to-port="1" />
-		<edge from-layer="1872" from-port="2" to-layer="1906" to-port="1" />
-		<edge from-layer="1873" from-port="2" to-layer="1874" to-port="1" />
-		<edge from-layer="1874" from-port="2" to-layer="1879" to-port="0" />
-		<edge from-layer="1875" from-port="0" to-layer="1878" to-port="1" />
-		<edge from-layer="1876" from-port="0" to-layer="1878" to-port="2" />
-		<edge from-layer="1877" from-port="0" to-layer="1878" to-port="3" />
-		<edge from-layer="1878" from-port="4" to-layer="1879" to-port="1" />
-		<edge from-layer="1879" from-port="2" to-layer="1970" to-port="0" />
-		<edge from-layer="1879" from-port="2" to-layer="1937" to-port="0" />
-		<edge from-layer="1880" from-port="0" to-layer="1883" to-port="1" />
-		<edge from-layer="1881" from-port="0" to-layer="1883" to-port="2" />
-		<edge from-layer="1882" from-port="0" to-layer="1883" to-port="3" />
-		<edge from-layer="1883" from-port="4" to-layer="1898" to-port="0" />
-		<edge from-layer="1883" from-port="4" to-layer="1904" to-port="0" />
-		<edge from-layer="1883" from-port="4" to-layer="1887" to-port="0" />
-		<edge from-layer="1883" from-port="4" to-layer="1884" to-port="0" />
-		<edge from-layer="1884" from-port="2" to-layer="1907" to-port="0" />
-		<edge from-layer="1885" from-port="0" to-layer="1895" to-port="0" />
-		<edge from-layer="1886" from-port="0" to-layer="1895" to-port="1" />
-		<edge from-layer="1887" from-port="1" to-layer="1889" to-port="0" />
-		<edge from-layer="1888" from-port="0" to-layer="1889" to-port="2" />
-		<edge from-layer="1889" from-port="3" to-layer="1890" to-port="0" />
-		<edge from-layer="1890" from-port="2" to-layer="1891" to-port="0" />
-		<edge from-layer="1891" from-port="1" to-layer="1893" to-port="0" />
-		<edge from-layer="1892" from-port="0" to-layer="1893" to-port="1" />
-		<edge from-layer="1893" from-port="2" to-layer="1895" to-port="2" />
-		<edge from-layer="1894" from-port="0" to-layer="1895" to-port="3" />
-		<edge from-layer="1895" from-port="4" to-layer="1898" to-port="1" />
-		<edge from-layer="1895" from-port="4" to-layer="1904" to-port="2" />
-		<edge from-layer="1896" from-port="0" to-layer="1898" to-port="2" />
-		<edge from-layer="1897" from-port="0" to-layer="1898" to-port="3" />
-		<edge from-layer="1898" from-port="4" to-layer="1901" to-port="0" />
-		<edge from-layer="1899" from-port="0" to-layer="1900" to-port="0" />
-		<edge from-layer="1900" from-port="1" to-layer="1901" to-port="1" />
-		<edge from-layer="1901" from-port="2" to-layer="1905" to-port="0" />
-		<edge from-layer="1902" from-port="0" to-layer="1904" to-port="1" />
-		<edge from-layer="1903" from-port="0" to-layer="1904" to-port="3" />
-		<edge from-layer="1904" from-port="4" to-layer="1905" to-port="1" />
-		<edge from-layer="1905" from-port="2" to-layer="1906" to-port="0" />
-		<edge from-layer="1906" from-port="2" to-layer="1907" to-port="1" />
-		<edge from-layer="1907" from-port="2" to-layer="1912" to-port="0" />
-		<edge from-layer="1908" from-port="0" to-layer="1911" to-port="1" />
-		<edge from-layer="1909" from-port="0" to-layer="1911" to-port="2" />
-		<edge from-layer="1910" from-port="0" to-layer="1911" to-port="3" />
-		<edge from-layer="1911" from-port="4" to-layer="1912" to-port="1" />
-		<edge from-layer="1912" from-port="2" to-layer="1913" to-port="1" />
-		<edge from-layer="1913" from-port="2" to-layer="1934" to-port="0" />
-		<edge from-layer="1913" from-port="2" to-layer="2340" to-port="0" />
-		<edge from-layer="1913" from-port="2" to-layer="1970" to-port="1" />
-		<edge from-layer="1914" from-port="0" to-layer="1915" to-port="0" />
-		<edge from-layer="1915" from-port="1" to-layer="1920" to-port="0" />
-		<edge from-layer="1916" from-port="0" to-layer="1919" to-port="1" />
-		<edge from-layer="1917" from-port="0" to-layer="1919" to-port="2" />
-		<edge from-layer="1918" from-port="0" to-layer="1919" to-port="3" />
-		<edge from-layer="1919" from-port="4" to-layer="1920" to-port="1" />
-		<edge from-layer="1920" from-port="2" to-layer="1921" to-port="0" />
-		<edge from-layer="1921" from-port="1" to-layer="1923" to-port="0" />
-		<edge from-layer="1922" from-port="0" to-layer="1923" to-port="2" />
-		<edge from-layer="1923" from-port="3" to-layer="1930" to-port="0" />
-		<edge from-layer="1924" from-port="0" to-layer="1927" to-port="1" />
-		<edge from-layer="1925" from-port="0" to-layer="1927" to-port="2" />
-		<edge from-layer="1926" from-port="0" to-layer="1927" to-port="3" />
-		<edge from-layer="1927" from-port="4" to-layer="1929" to-port="0" />
-		<edge from-layer="1928" from-port="0" to-layer="1929" to-port="1" />
-		<edge from-layer="1929" from-port="2" to-layer="1930" to-port="1" />
-		<edge from-layer="1930" from-port="2" to-layer="2339" to-port="0" />
-		<edge from-layer="1930" from-port="2" to-layer="1970" to-port="2" />
-		<edge from-layer="1931" from-port="0" to-layer="1949" to-port="0" />
-		<edge from-layer="1932" from-port="0" to-layer="1944" to-port="0" />
-		<edge from-layer="1933" from-port="0" to-layer="1947" to-port="1" />
-		<edge from-layer="1933" from-port="0" to-layer="1944" to-port="1" />
-		<edge from-layer="1934" from-port="1" to-layer="1936" to-port="0" />
-		<edge from-layer="1935" from-port="0" to-layer="1936" to-port="2" />
-		<edge from-layer="1936" from-port="3" to-layer="1940" to-port="0" />
-		<edge from-layer="1936" from-port="3" to-layer="1946" to-port="0" />
-		<edge from-layer="1937" from-port="1" to-layer="1965" to-port="0" />
-		<edge from-layer="1937" from-port="1" to-layer="1939" to-port="0" />
-		<edge from-layer="1938" from-port="0" to-layer="1939" to-port="2" />
-		<edge from-layer="1939" from-port="3" to-layer="1940" to-port="1" />
-		<edge from-layer="1940" from-port="2" to-layer="1942" to-port="0" />
-		<edge from-layer="1941" from-port="0" to-layer="1946" to-port="1" />
-		<edge from-layer="1941" from-port="0" to-layer="1942" to-port="1" />
-		<edge from-layer="1942" from-port="2" to-layer="1944" to-port="2" />
-		<edge from-layer="1943" from-port="0" to-layer="1947" to-port="3" />
-		<edge from-layer="1943" from-port="0" to-layer="1944" to-port="3" />
-		<edge from-layer="1944" from-port="4" to-layer="1949" to-port="1" />
-		<edge from-layer="1945" from-port="0" to-layer="1947" to-port="0" />
-		<edge from-layer="1946" from-port="2" to-layer="1954" to-port="2" />
-		<edge from-layer="1946" from-port="2" to-layer="1947" to-port="2" />
-		<edge from-layer="1947" from-port="4" to-layer="1949" to-port="2" />
-		<edge from-layer="1948" from-port="0" to-layer="1949" to-port="3" />
-		<edge from-layer="1949" from-port="4" to-layer="1956" to-port="0" />
-		<edge from-layer="1950" from-port="0" to-layer="1956" to-port="1" />
-		<edge from-layer="1951" from-port="0" to-layer="1954" to-port="0" />
-		<edge from-layer="1952" from-port="0" to-layer="1954" to-port="1" />
-		<edge from-layer="1953" from-port="0" to-layer="1954" to-port="3" />
-		<edge from-layer="1954" from-port="4" to-layer="1956" to-port="2" />
-		<edge from-layer="1955" from-port="0" to-layer="1956" to-port="3" />
-		<edge from-layer="1956" from-port="4" to-layer="1960" to-port="0" />
-		<edge from-layer="1957" from-port="0" to-layer="1958" to-port="0" />
-		<edge from-layer="1958" from-port="1" to-layer="1960" to-port="1" />
-		<edge from-layer="1959" from-port="0" to-layer="1960" to-port="2" />
-		<edge from-layer="1960" from-port="3" to-layer="1961" to-port="0" />
-		<edge from-layer="1961" from-port="2" to-layer="1969" to-port="0" />
-		<edge from-layer="1961" from-port="2" to-layer="1962" to-port="0" />
-		<edge from-layer="1962" from-port="1" to-layer="1968" to-port="0" />
-		<edge from-layer="1963" from-port="0" to-layer="1965" to-port="1" />
-		<edge from-layer="1964" from-port="0" to-layer="1965" to-port="2" />
-		<edge from-layer="1965" from-port="3" to-layer="1966" to-port="0" />
-		<edge from-layer="1966" from-port="4" to-layer="1967" to-port="0" />
-		<edge from-layer="1967" from-port="1" to-layer="1968" to-port="1" />
-		<edge from-layer="1968" from-port="2" to-layer="1969" to-port="1" />
-		<edge from-layer="1969" from-port="2" to-layer="1970" to-port="3" />
-		<edge from-layer="1970" from-port="4" to-layer="1972" to-port="0" />
-		<edge from-layer="1971" from-port="0" to-layer="1972" to-port="1" />
-		<edge from-layer="1972" from-port="2" to-layer="1974" to-port="0" />
+		<edge from-layer="1767" from-port="1" to-layer="1768" to-port="1" />
+		<edge from-layer="1768" from-port="2" to-layer="1770" to-port="0" />
+		<edge from-layer="1769" from-port="0" to-layer="1770" to-port="1" />
+		<edge from-layer="1770" from-port="2" to-layer="1772" to-port="0" />
+		<edge from-layer="1771" from-port="0" to-layer="1772" to-port="1" />
+		<edge from-layer="1772" from-port="2" to-layer="1853" to-port="0" />
+		<edge from-layer="1772" from-port="2" to-layer="1814" to-port="0" />
+		<edge from-layer="1772" from-port="2" to-layer="1776" to-port="0" />
+		<edge from-layer="1773" from-port="0" to-layer="1776" to-port="1" />
+		<edge from-layer="1774" from-port="0" to-layer="1776" to-port="2" />
+		<edge from-layer="1775" from-port="0" to-layer="1776" to-port="3" />
+		<edge from-layer="1776" from-port="5" to-layer="1778" to-port="0" />
+		<edge from-layer="1777" from-port="0" to-layer="1778" to-port="1" />
+		<edge from-layer="1778" from-port="2" to-layer="1800" to-port="0" />
+		<edge from-layer="1778" from-port="2" to-layer="1782" to-port="0" />
+		<edge from-layer="1779" from-port="0" to-layer="1782" to-port="1" />
+		<edge from-layer="1780" from-port="0" to-layer="1782" to-port="2" />
+		<edge from-layer="1781" from-port="0" to-layer="1782" to-port="3" />
+		<edge from-layer="1782" from-port="5" to-layer="1793" to-port="0" />
+		<edge from-layer="1782" from-port="5" to-layer="1783" to-port="0" />
+		<edge from-layer="1782" from-port="5" to-layer="1787" to-port="0" />
+		<edge from-layer="1783" from-port="2" to-layer="1796" to-port="0" />
+		<edge from-layer="1784" from-port="0" to-layer="1787" to-port="1" />
+		<edge from-layer="1785" from-port="0" to-layer="1787" to-port="2" />
+		<edge from-layer="1786" from-port="0" to-layer="1787" to-port="3" />
+		<edge from-layer="1787" from-port="5" to-layer="1789" to-port="0" />
+		<edge from-layer="1788" from-port="0" to-layer="1789" to-port="1" />
+		<edge from-layer="1789" from-port="2" to-layer="1794" to-port="0" />
+		<edge from-layer="1790" from-port="0" to-layer="1793" to-port="1" />
+		<edge from-layer="1791" from-port="0" to-layer="1793" to-port="2" />
+		<edge from-layer="1792" from-port="0" to-layer="1793" to-port="3" />
+		<edge from-layer="1793" from-port="5" to-layer="1794" to-port="1" />
+		<edge from-layer="1794" from-port="2" to-layer="1795" to-port="0" />
+		<edge from-layer="1795" from-port="2" to-layer="1796" to-port="1" />
+		<edge from-layer="1796" from-port="2" to-layer="1801" to-port="0" />
+		<edge from-layer="1797" from-port="0" to-layer="1800" to-port="1" />
+		<edge from-layer="1798" from-port="0" to-layer="1800" to-port="2" />
+		<edge from-layer="1799" from-port="0" to-layer="1800" to-port="3" />
+		<edge from-layer="1800" from-port="5" to-layer="1801" to-port="1" />
+		<edge from-layer="1801" from-port="2" to-layer="1865" to-port="0" />
+		<edge from-layer="1802" from-port="0" to-layer="1807" to-port="0" />
+		<edge from-layer="1803" from-port="0" to-layer="1806" to-port="1" />
+		<edge from-layer="1804" from-port="0" to-layer="1806" to-port="2" />
+		<edge from-layer="1805" from-port="0" to-layer="1806" to-port="3" />
+		<edge from-layer="1806" from-port="4" to-layer="1807" to-port="1" />
+		<edge from-layer="1807" from-port="2" to-layer="1808" to-port="0" />
+		<edge from-layer="1808" from-port="1" to-layer="1810" to-port="0" />
+		<edge from-layer="1809" from-port="0" to-layer="1810" to-port="2" />
+		<edge from-layer="1810" from-port="3" to-layer="1858" to-port="0" />
+		<edge from-layer="1810" from-port="3" to-layer="1840" to-port="0" />
+		<edge from-layer="1811" from-port="0" to-layer="1814" to-port="1" />
+		<edge from-layer="1812" from-port="0" to-layer="1814" to-port="2" />
+		<edge from-layer="1813" from-port="0" to-layer="1814" to-port="3" />
+		<edge from-layer="1814" from-port="5" to-layer="1816" to-port="0" />
+		<edge from-layer="1815" from-port="0" to-layer="1816" to-port="1" />
+		<edge from-layer="1816" from-port="2" to-layer="1820" to-port="0" />
+		<edge from-layer="1816" from-port="2" to-layer="1838" to-port="0" />
+		<edge from-layer="1817" from-port="0" to-layer="1820" to-port="1" />
+		<edge from-layer="1818" from-port="0" to-layer="1820" to-port="2" />
+		<edge from-layer="1819" from-port="0" to-layer="1820" to-port="3" />
+		<edge from-layer="1820" from-port="5" to-layer="1831" to-port="0" />
+		<edge from-layer="1820" from-port="5" to-layer="1825" to-port="0" />
+		<edge from-layer="1820" from-port="5" to-layer="1821" to-port="0" />
+		<edge from-layer="1821" from-port="2" to-layer="1834" to-port="0" />
+		<edge from-layer="1822" from-port="0" to-layer="1825" to-port="1" />
+		<edge from-layer="1823" from-port="0" to-layer="1825" to-port="2" />
+		<edge from-layer="1824" from-port="0" to-layer="1825" to-port="3" />
+		<edge from-layer="1825" from-port="5" to-layer="1827" to-port="0" />
+		<edge from-layer="1826" from-port="0" to-layer="1827" to-port="1" />
+		<edge from-layer="1827" from-port="2" to-layer="1832" to-port="0" />
+		<edge from-layer="1828" from-port="0" to-layer="1831" to-port="1" />
+		<edge from-layer="1829" from-port="0" to-layer="1831" to-port="2" />
+		<edge from-layer="1830" from-port="0" to-layer="1831" to-port="3" />
+		<edge from-layer="1831" from-port="5" to-layer="1832" to-port="1" />
+		<edge from-layer="1832" from-port="2" to-layer="1833" to-port="0" />
+		<edge from-layer="1833" from-port="2" to-layer="1834" to-port="1" />
+		<edge from-layer="1834" from-port="2" to-layer="1839" to-port="0" />
+		<edge from-layer="1835" from-port="0" to-layer="1838" to-port="1" />
+		<edge from-layer="1836" from-port="0" to-layer="1838" to-port="2" />
+		<edge from-layer="1837" from-port="0" to-layer="1838" to-port="3" />
+		<edge from-layer="1838" from-port="5" to-layer="1839" to-port="1" />
+		<edge from-layer="1839" from-port="2" to-layer="1840" to-port="1" />
+		<edge from-layer="1840" from-port="2" to-layer="1865" to-port="1" />
+		<edge from-layer="1840" from-port="2" to-layer="5155" to-port="0" />
+		<edge from-layer="1841" from-port="0" to-layer="1846" to-port="0" />
+		<edge from-layer="1842" from-port="0" to-layer="1845" to-port="1" />
+		<edge from-layer="1843" from-port="0" to-layer="1845" to-port="2" />
+		<edge from-layer="1844" from-port="0" to-layer="1845" to-port="3" />
+		<edge from-layer="1845" from-port="4" to-layer="1846" to-port="1" />
+		<edge from-layer="1846" from-port="2" to-layer="1847" to-port="0" />
+		<edge from-layer="1847" from-port="1" to-layer="1849" to-port="0" />
+		<edge from-layer="1848" from-port="0" to-layer="1849" to-port="2" />
+		<edge from-layer="1849" from-port="3" to-layer="1856" to-port="0" />
+		<edge from-layer="1850" from-port="0" to-layer="1853" to-port="1" />
+		<edge from-layer="1851" from-port="0" to-layer="1853" to-port="2" />
+		<edge from-layer="1852" from-port="0" to-layer="1853" to-port="3" />
+		<edge from-layer="1853" from-port="5" to-layer="1855" to-port="0" />
+		<edge from-layer="1854" from-port="0" to-layer="1855" to-port="1" />
+		<edge from-layer="1855" from-port="2" to-layer="1856" to-port="1" />
+		<edge from-layer="1856" from-port="2" to-layer="5154" to-port="0" />
+		<edge from-layer="1856" from-port="2" to-layer="1865" to-port="2" />
+		<edge from-layer="1857" from-port="0" to-layer="1864" to-port="1" />
+		<edge from-layer="1858" from-port="1" to-layer="1861" to-port="0" />
+		<edge from-layer="1859" from-port="0" to-layer="1861" to-port="1" />
+		<edge from-layer="1860" from-port="0" to-layer="1861" to-port="2" />
+		<edge from-layer="1861" from-port="3" to-layer="1862" to-port="0" />
+		<edge from-layer="1862" from-port="2" to-layer="1864" to-port="2" />
+		<edge from-layer="1863" from-port="0" to-layer="1864" to-port="3" />
+		<edge from-layer="1864" from-port="5" to-layer="1865" to-port="3" />
+		<edge from-layer="1865" from-port="4" to-layer="1867" to-port="0" />
+		<edge from-layer="1866" from-port="0" to-layer="1867" to-port="1" />
+		<edge from-layer="1867" from-port="2" to-layer="1869" to-port="0" />
+		<edge from-layer="1868" from-port="0" to-layer="1869" to-port="1" />
+		<edge from-layer="1869" from-port="2" to-layer="1878" to-port="0" />
+		<edge from-layer="1870" from-port="0" to-layer="1871" to-port="0" />
+		<edge from-layer="1871" from-port="1" to-layer="1874" to-port="0" />
+		<edge from-layer="1872" from-port="0" to-layer="1873" to-port="0" />
+		<edge from-layer="1873" from-port="1" to-layer="1874" to-port="1" />
+		<edge from-layer="1874" from-port="2" to-layer="1876" to-port="0" />
+		<edge from-layer="1875" from-port="0" to-layer="1876" to-port="1" />
+		<edge from-layer="1876" from-port="2" to-layer="1877" to-port="0" />
+		<edge from-layer="1877" from-port="1" to-layer="1878" to-port="1" />
+		<edge from-layer="1878" from-port="2" to-layer="1880" to-port="0" />
+		<edge from-layer="1879" from-port="0" to-layer="1880" to-port="1" />
+		<edge from-layer="1880" from-port="2" to-layer="1881" to-port="1" />
+		<edge from-layer="1881" from-port="2" to-layer="1882" to-port="0" />
+		<edge from-layer="1882" from-port="2" to-layer="1884" to-port="0" />
+		<edge from-layer="1882" from-port="2" to-layer="2040" to-port="1" />
+		<edge from-layer="1883" from-port="0" to-layer="1884" to-port="1" />
+		<edge from-layer="1884" from-port="2" to-layer="1886" to-port="0" />
+		<edge from-layer="1884" from-port="2" to-layer="1915" to-port="0" />
+		<edge from-layer="1885" from-port="0" to-layer="1886" to-port="1" />
+		<edge from-layer="1886" from-port="2" to-layer="1888" to-port="0" />
+		<edge from-layer="1887" from-port="0" to-layer="1888" to-port="1" />
+		<edge from-layer="1888" from-port="2" to-layer="1897" to-port="0" />
+		<edge from-layer="1889" from-port="0" to-layer="1890" to-port="0" />
+		<edge from-layer="1890" from-port="1" to-layer="1893" to-port="0" />
+		<edge from-layer="1891" from-port="0" to-layer="1892" to-port="0" />
+		<edge from-layer="1892" from-port="1" to-layer="1893" to-port="1" />
+		<edge from-layer="1893" from-port="2" to-layer="1895" to-port="0" />
+		<edge from-layer="1894" from-port="0" to-layer="1895" to-port="1" />
+		<edge from-layer="1895" from-port="2" to-layer="1896" to-port="0" />
+		<edge from-layer="1896" from-port="1" to-layer="1897" to-port="1" />
+		<edge from-layer="1897" from-port="2" to-layer="1900" to-port="0" />
+		<edge from-layer="1898" from-port="0" to-layer="1899" to-port="0" />
+		<edge from-layer="1899" from-port="1" to-layer="1900" to-port="1" />
+		<edge from-layer="1900" from-port="2" to-layer="1901" to-port="0" />
+		<edge from-layer="1901" from-port="1" to-layer="1910" to-port="0" />
+		<edge from-layer="1902" from-port="0" to-layer="1903" to-port="0" />
+		<edge from-layer="1903" from-port="1" to-layer="1906" to-port="0" />
+		<edge from-layer="1904" from-port="0" to-layer="1905" to-port="0" />
+		<edge from-layer="1905" from-port="1" to-layer="1906" to-port="1" />
+		<edge from-layer="1906" from-port="2" to-layer="1908" to-port="0" />
+		<edge from-layer="1907" from-port="0" to-layer="1908" to-port="1" />
+		<edge from-layer="1908" from-port="2" to-layer="1909" to-port="0" />
+		<edge from-layer="1909" from-port="1" to-layer="1910" to-port="1" />
+		<edge from-layer="1910" from-port="2" to-layer="1913" to-port="0" />
+		<edge from-layer="1911" from-port="0" to-layer="1912" to-port="0" />
+		<edge from-layer="1912" from-port="1" to-layer="1913" to-port="1" />
+		<edge from-layer="1913" from-port="2" to-layer="2039" to-port="0" />
+		<edge from-layer="1914" from-port="0" to-layer="1915" to-port="1" />
+		<edge from-layer="1915" from-port="2" to-layer="1917" to-port="0" />
+		<edge from-layer="1916" from-port="0" to-layer="1917" to-port="1" />
+		<edge from-layer="1917" from-port="2" to-layer="1926" to-port="0" />
+		<edge from-layer="1918" from-port="0" to-layer="1919" to-port="0" />
+		<edge from-layer="1919" from-port="1" to-layer="1922" to-port="0" />
+		<edge from-layer="1920" from-port="0" to-layer="1921" to-port="0" />
+		<edge from-layer="1921" from-port="1" to-layer="1922" to-port="1" />
+		<edge from-layer="1922" from-port="2" to-layer="1924" to-port="0" />
+		<edge from-layer="1923" from-port="0" to-layer="1924" to-port="1" />
+		<edge from-layer="1924" from-port="2" to-layer="1925" to-port="0" />
+		<edge from-layer="1925" from-port="1" to-layer="1926" to-port="1" />
+		<edge from-layer="1926" from-port="2" to-layer="1928" to-port="0" />
+		<edge from-layer="1927" from-port="0" to-layer="1928" to-port="1" />
+		<edge from-layer="1928" from-port="2" to-layer="1930" to-port="0" />
+		<edge from-layer="1929" from-port="0" to-layer="1930" to-port="1" />
+		<edge from-layer="1930" from-port="2" to-layer="2011" to-port="0" />
+		<edge from-layer="1930" from-port="2" to-layer="1972" to-port="0" />
+		<edge from-layer="1930" from-port="2" to-layer="1934" to-port="0" />
+		<edge from-layer="1931" from-port="0" to-layer="1934" to-port="1" />
+		<edge from-layer="1932" from-port="0" to-layer="1934" to-port="2" />
+		<edge from-layer="1933" from-port="0" to-layer="1934" to-port="3" />
+		<edge from-layer="1934" from-port="5" to-layer="1936" to-port="0" />
+		<edge from-layer="1935" from-port="0" to-layer="1936" to-port="1" />
+		<edge from-layer="1936" from-port="2" to-layer="1958" to-port="0" />
+		<edge from-layer="1936" from-port="2" to-layer="1940" to-port="0" />
+		<edge from-layer="1937" from-port="0" to-layer="1940" to-port="1" />
+		<edge from-layer="1938" from-port="0" to-layer="1940" to-port="2" />
+		<edge from-layer="1939" from-port="0" to-layer="1940" to-port="3" />
+		<edge from-layer="1940" from-port="5" to-layer="1945" to-port="0" />
+		<edge from-layer="1940" from-port="5" to-layer="1951" to-port="0" />
+		<edge from-layer="1940" from-port="5" to-layer="1941" to-port="0" />
+		<edge from-layer="1941" from-port="2" to-layer="1954" to-port="0" />
+		<edge from-layer="1942" from-port="0" to-layer="1945" to-port="1" />
+		<edge from-layer="1943" from-port="0" to-layer="1945" to-port="2" />
+		<edge from-layer="1944" from-port="0" to-layer="1945" to-port="3" />
+		<edge from-layer="1945" from-port="5" to-layer="1947" to-port="0" />
+		<edge from-layer="1946" from-port="0" to-layer="1947" to-port="1" />
+		<edge from-layer="1947" from-port="2" to-layer="1952" to-port="0" />
+		<edge from-layer="1948" from-port="0" to-layer="1951" to-port="1" />
+		<edge from-layer="1949" from-port="0" to-layer="1951" to-port="2" />
+		<edge from-layer="1950" from-port="0" to-layer="1951" to-port="3" />
+		<edge from-layer="1951" from-port="5" to-layer="1952" to-port="1" />
+		<edge from-layer="1952" from-port="2" to-layer="1953" to-port="0" />
+		<edge from-layer="1953" from-port="2" to-layer="1954" to-port="1" />
+		<edge from-layer="1954" from-port="2" to-layer="1959" to-port="0" />
+		<edge from-layer="1955" from-port="0" to-layer="1958" to-port="1" />
+		<edge from-layer="1956" from-port="0" to-layer="1958" to-port="2" />
+		<edge from-layer="1957" from-port="0" to-layer="1958" to-port="3" />
+		<edge from-layer="1958" from-port="5" to-layer="1959" to-port="1" />
+		<edge from-layer="1959" from-port="2" to-layer="2023" to-port="0" />
+		<edge from-layer="1960" from-port="0" to-layer="1965" to-port="0" />
+		<edge from-layer="1961" from-port="0" to-layer="1964" to-port="1" />
+		<edge from-layer="1962" from-port="0" to-layer="1964" to-port="2" />
+		<edge from-layer="1963" from-port="0" to-layer="1964" to-port="3" />
+		<edge from-layer="1964" from-port="4" to-layer="1965" to-port="1" />
+		<edge from-layer="1965" from-port="2" to-layer="1966" to-port="0" />
+		<edge from-layer="1966" from-port="1" to-layer="1968" to-port="0" />
+		<edge from-layer="1967" from-port="0" to-layer="1968" to-port="2" />
+		<edge from-layer="1968" from-port="3" to-layer="2016" to-port="0" />
+		<edge from-layer="1968" from-port="3" to-layer="1998" to-port="0" />
+		<edge from-layer="1969" from-port="0" to-layer="1972" to-port="1" />
+		<edge from-layer="1970" from-port="0" to-layer="1972" to-port="2" />
+		<edge from-layer="1971" from-port="0" to-layer="1972" to-port="3" />
+		<edge from-layer="1972" from-port="5" to-layer="1974" to-port="0" />
 		<edge from-layer="1973" from-port="0" to-layer="1974" to-port="1" />
-		<edge from-layer="1974" from-port="2" to-layer="1983" to-port="0" />
-		<edge from-layer="1975" from-port="0" to-layer="1976" to-port="0" />
-		<edge from-layer="1976" from-port="1" to-layer="1979" to-port="0" />
-		<edge from-layer="1977" from-port="0" to-layer="1978" to-port="0" />
-		<edge from-layer="1978" from-port="1" to-layer="1979" to-port="1" />
-		<edge from-layer="1979" from-port="2" to-layer="1981" to-port="0" />
-		<edge from-layer="1980" from-port="0" to-layer="1981" to-port="1" />
-		<edge from-layer="1981" from-port="2" to-layer="1982" to-port="0" />
-		<edge from-layer="1982" from-port="1" to-layer="1983" to-port="1" />
-		<edge from-layer="1983" from-port="2" to-layer="1986" to-port="0" />
-		<edge from-layer="1984" from-port="0" to-layer="1985" to-port="0" />
-		<edge from-layer="1985" from-port="1" to-layer="1986" to-port="1" />
-		<edge from-layer="1986" from-port="2" to-layer="1987" to-port="1" />
-		<edge from-layer="1987" from-port="2" to-layer="1988" to-port="0" />
-		<edge from-layer="1988" from-port="2" to-layer="2230" to-port="1" />
-		<edge from-layer="1988" from-port="2" to-layer="1990" to-port="0" />
-		<edge from-layer="1989" from-port="0" to-layer="1990" to-port="1" />
-		<edge from-layer="1990" from-port="2" to-layer="1993" to-port="0" />
-		<edge from-layer="1990" from-port="2" to-layer="2024" to-port="0" />
-		<edge from-layer="1991" from-port="0" to-layer="1992" to-port="0" />
-		<edge from-layer="1992" from-port="1" to-layer="1993" to-port="1" />
-		<edge from-layer="1993" from-port="2" to-layer="1996" to-port="0" />
-		<edge from-layer="1994" from-port="0" to-layer="1995" to-port="0" />
-		<edge from-layer="1995" from-port="1" to-layer="1996" to-port="1" />
-		<edge from-layer="1996" from-port="2" to-layer="2005" to-port="0" />
-		<edge from-layer="1997" from-port="0" to-layer="1998" to-port="0" />
-		<edge from-layer="1998" from-port="1" to-layer="2001" to-port="0" />
-		<edge from-layer="1999" from-port="0" to-layer="2000" to-port="0" />
-		<edge from-layer="2000" from-port="1" to-layer="2001" to-port="1" />
-		<edge from-layer="2001" from-port="2" to-layer="2003" to-port="0" />
-		<edge from-layer="2002" from-port="0" to-layer="2003" to-port="1" />
-		<edge from-layer="2003" from-port="2" to-layer="2004" to-port="0" />
-		<edge from-layer="2004" from-port="1" to-layer="2005" to-port="1" />
-		<edge from-layer="2005" from-port="2" to-layer="2008" to-port="0" />
-		<edge from-layer="2006" from-port="0" to-layer="2007" to-port="0" />
-		<edge from-layer="2007" from-port="1" to-layer="2008" to-port="1" />
-		<edge from-layer="2008" from-port="2" to-layer="2009" to-port="0" />
-		<edge from-layer="2009" from-port="1" to-layer="2018" to-port="0" />
-		<edge from-layer="2010" from-port="0" to-layer="2011" to-port="0" />
-		<edge from-layer="2011" from-port="1" to-layer="2014" to-port="0" />
-		<edge from-layer="2012" from-port="0" to-layer="2013" to-port="0" />
-		<edge from-layer="2013" from-port="1" to-layer="2014" to-port="1" />
-		<edge from-layer="2014" from-port="2" to-layer="2016" to-port="0" />
-		<edge from-layer="2015" from-port="0" to-layer="2016" to-port="1" />
-		<edge from-layer="2016" from-port="2" to-layer="2017" to-port="0" />
-		<edge from-layer="2017" from-port="1" to-layer="2018" to-port="1" />
-		<edge from-layer="2018" from-port="2" to-layer="2021" to-port="0" />
-		<edge from-layer="2019" from-port="0" to-layer="2020" to-port="0" />
-		<edge from-layer="2020" from-port="1" to-layer="2021" to-port="1" />
-		<edge from-layer="2021" from-port="2" to-layer="2229" to-port="0" />
-		<edge from-layer="2022" from-port="0" to-layer="2023" to-port="0" />
-		<edge from-layer="2023" from-port="1" to-layer="2024" to-port="1" />
-		<edge from-layer="2024" from-port="2" to-layer="2027" to-port="0" />
-		<edge from-layer="2025" from-port="0" to-layer="2026" to-port="0" />
-		<edge from-layer="2026" from-port="1" to-layer="2027" to-port="1" />
+		<edge from-layer="1974" from-port="2" to-layer="1978" to-port="0" />
+		<edge from-layer="1974" from-port="2" to-layer="1996" to-port="0" />
+		<edge from-layer="1975" from-port="0" to-layer="1978" to-port="1" />
+		<edge from-layer="1976" from-port="0" to-layer="1978" to-port="2" />
+		<edge from-layer="1977" from-port="0" to-layer="1978" to-port="3" />
+		<edge from-layer="1978" from-port="5" to-layer="1979" to-port="0" />
+		<edge from-layer="1978" from-port="5" to-layer="1983" to-port="0" />
+		<edge from-layer="1978" from-port="5" to-layer="1989" to-port="0" />
+		<edge from-layer="1979" from-port="2" to-layer="1992" to-port="0" />
+		<edge from-layer="1980" from-port="0" to-layer="1983" to-port="1" />
+		<edge from-layer="1981" from-port="0" to-layer="1983" to-port="2" />
+		<edge from-layer="1982" from-port="0" to-layer="1983" to-port="3" />
+		<edge from-layer="1983" from-port="5" to-layer="1985" to-port="0" />
+		<edge from-layer="1984" from-port="0" to-layer="1985" to-port="1" />
+		<edge from-layer="1985" from-port="2" to-layer="1990" to-port="0" />
+		<edge from-layer="1986" from-port="0" to-layer="1989" to-port="1" />
+		<edge from-layer="1987" from-port="0" to-layer="1989" to-port="2" />
+		<edge from-layer="1988" from-port="0" to-layer="1989" to-port="3" />
+		<edge from-layer="1989" from-port="5" to-layer="1990" to-port="1" />
+		<edge from-layer="1990" from-port="2" to-layer="1991" to-port="0" />
+		<edge from-layer="1991" from-port="2" to-layer="1992" to-port="1" />
+		<edge from-layer="1992" from-port="2" to-layer="1997" to-port="0" />
+		<edge from-layer="1993" from-port="0" to-layer="1996" to-port="1" />
+		<edge from-layer="1994" from-port="0" to-layer="1996" to-port="2" />
+		<edge from-layer="1995" from-port="0" to-layer="1996" to-port="3" />
+		<edge from-layer="1996" from-port="5" to-layer="1997" to-port="1" />
+		<edge from-layer="1997" from-port="2" to-layer="1998" to-port="1" />
+		<edge from-layer="1998" from-port="2" to-layer="2023" to-port="1" />
+		<edge from-layer="1998" from-port="2" to-layer="5153" to-port="0" />
+		<edge from-layer="1999" from-port="0" to-layer="2004" to-port="0" />
+		<edge from-layer="2000" from-port="0" to-layer="2003" to-port="1" />
+		<edge from-layer="2001" from-port="0" to-layer="2003" to-port="2" />
+		<edge from-layer="2002" from-port="0" to-layer="2003" to-port="3" />
+		<edge from-layer="2003" from-port="4" to-layer="2004" to-port="1" />
+		<edge from-layer="2004" from-port="2" to-layer="2005" to-port="0" />
+		<edge from-layer="2005" from-port="1" to-layer="2007" to-port="0" />
+		<edge from-layer="2006" from-port="0" to-layer="2007" to-port="2" />
+		<edge from-layer="2007" from-port="3" to-layer="2014" to-port="0" />
+		<edge from-layer="2008" from-port="0" to-layer="2011" to-port="1" />
+		<edge from-layer="2009" from-port="0" to-layer="2011" to-port="2" />
+		<edge from-layer="2010" from-port="0" to-layer="2011" to-port="3" />
+		<edge from-layer="2011" from-port="5" to-layer="2013" to-port="0" />
+		<edge from-layer="2012" from-port="0" to-layer="2013" to-port="1" />
+		<edge from-layer="2013" from-port="2" to-layer="2014" to-port="1" />
+		<edge from-layer="2014" from-port="2" to-layer="2023" to-port="2" />
+		<edge from-layer="2014" from-port="2" to-layer="5152" to-port="0" />
+		<edge from-layer="2015" from-port="0" to-layer="2022" to-port="1" />
+		<edge from-layer="2016" from-port="1" to-layer="2019" to-port="0" />
+		<edge from-layer="2017" from-port="0" to-layer="2019" to-port="1" />
+		<edge from-layer="2018" from-port="0" to-layer="2019" to-port="2" />
+		<edge from-layer="2019" from-port="3" to-layer="2020" to-port="0" />
+		<edge from-layer="2020" from-port="2" to-layer="2022" to-port="2" />
+		<edge from-layer="2021" from-port="0" to-layer="2022" to-port="3" />
+		<edge from-layer="2022" from-port="5" to-layer="2023" to-port="3" />
+		<edge from-layer="2023" from-port="4" to-layer="2025" to-port="0" />
+		<edge from-layer="2024" from-port="0" to-layer="2025" to-port="1" />
+		<edge from-layer="2025" from-port="2" to-layer="2027" to-port="0" />
+		<edge from-layer="2026" from-port="0" to-layer="2027" to-port="1" />
 		<edge from-layer="2027" from-port="2" to-layer="2036" to-port="0" />
 		<edge from-layer="2028" from-port="0" to-layer="2029" to-port="0" />
 		<edge from-layer="2029" from-port="1" to-layer="2032" to-port="0" />
@@ -123816,148 +83388,140 @@
 		<edge from-layer="2033" from-port="0" to-layer="2034" to-port="1" />
 		<edge from-layer="2034" from-port="2" to-layer="2035" to-port="0" />
 		<edge from-layer="2035" from-port="1" to-layer="2036" to-port="1" />
-		<edge from-layer="2036" from-port="2" to-layer="2039" to-port="0" />
-		<edge from-layer="2037" from-port="0" to-layer="2038" to-port="0" />
-		<edge from-layer="2038" from-port="1" to-layer="2039" to-port="1" />
-		<edge from-layer="2039" from-port="2" to-layer="2041" to-port="0" />
-		<edge from-layer="2040" from-port="0" to-layer="2041" to-port="1" />
-		<edge from-layer="2041" from-port="2" to-layer="2058" to-port="0" />
-		<edge from-layer="2041" from-port="2" to-layer="2169" to-port="0" />
-		<edge from-layer="2041" from-port="2" to-layer="2045" to-port="0" />
-		<edge from-layer="2042" from-port="0" to-layer="2045" to-port="1" />
-		<edge from-layer="2043" from-port="0" to-layer="2045" to-port="2" />
-		<edge from-layer="2044" from-port="0" to-layer="2045" to-port="3" />
-		<edge from-layer="2045" from-port="4" to-layer="2047" to-port="0" />
-		<edge from-layer="2046" from-port="0" to-layer="2047" to-port="1" />
-		<edge from-layer="2047" from-port="2" to-layer="2051" to-port="0" />
-		<edge from-layer="2047" from-port="2" to-layer="2120" to-port="0" />
-		<edge from-layer="2048" from-port="0" to-layer="2051" to-port="1" />
-		<edge from-layer="2049" from-port="0" to-layer="2051" to-port="2" />
-		<edge from-layer="2050" from-port="0" to-layer="2051" to-port="3" />
-		<edge from-layer="2051" from-port="4" to-layer="2099" to-port="0" />
-		<edge from-layer="2051" from-port="4" to-layer="2105" to-port="0" />
-		<edge from-layer="2051" from-port="4" to-layer="2088" to-port="0" />
-		<edge from-layer="2051" from-port="4" to-layer="2085" to-port="0" />
-		<edge from-layer="2052" from-port="0" to-layer="2053" to-port="0" />
-		<edge from-layer="2053" from-port="1" to-layer="2081" to-port="0" />
-		<edge from-layer="2054" from-port="0" to-layer="2081" to-port="1" />
-		<edge from-layer="2055" from-port="0" to-layer="2058" to-port="1" />
-		<edge from-layer="2056" from-port="0" to-layer="2058" to-port="2" />
-		<edge from-layer="2057" from-port="0" to-layer="2058" to-port="3" />
-		<edge from-layer="2058" from-port="4" to-layer="2060" to-port="0" />
-		<edge from-layer="2059" from-port="0" to-layer="2060" to-port="1" />
-		<edge from-layer="2060" from-port="2" to-layer="2153" to-port="0" />
-		<edge from-layer="2060" from-port="2" to-layer="2125" to-port="0" />
-		<edge from-layer="2060" from-port="2" to-layer="2061" to-port="0" />
-		<edge from-layer="2061" from-port="1" to-layer="2063" to-port="0" />
-		<edge from-layer="2062" from-port="0" to-layer="2063" to-port="2" />
-		<edge from-layer="2063" from-port="3" to-layer="2077" to-port="0" />
-		<edge from-layer="2064" from-port="0" to-layer="2065" to-port="0" />
-		<edge from-layer="2065" from-port="1" to-layer="2070" to-port="0" />
-		<edge from-layer="2066" from-port="0" to-layer="2069" to-port="1" />
-		<edge from-layer="2067" from-port="0" to-layer="2069" to-port="2" />
-		<edge from-layer="2068" from-port="0" to-layer="2069" to-port="3" />
-		<edge from-layer="2069" from-port="4" to-layer="2070" to-port="1" />
-		<edge from-layer="2070" from-port="2" to-layer="2071" to-port="0" />
-		<edge from-layer="2071" from-port="1" to-layer="2073" to-port="0" />
-		<edge from-layer="2072" from-port="0" to-layer="2073" to-port="2" />
-		<edge from-layer="2073" from-port="3" to-layer="2155" to-port="0" />
-		<edge from-layer="2073" from-port="3" to-layer="2074" to-port="0" />
-		<edge from-layer="2074" from-port="1" to-layer="2076" to-port="0" />
-		<edge from-layer="2075" from-port="0" to-layer="2076" to-port="2" />
-		<edge from-layer="2076" from-port="3" to-layer="2077" to-port="1" />
-		<edge from-layer="2077" from-port="2" to-layer="2079" to-port="0" />
-		<edge from-layer="2078" from-port="0" to-layer="2079" to-port="1" />
-		<edge from-layer="2079" from-port="2" to-layer="2111" to-port="2" />
-		<edge from-layer="2079" from-port="2" to-layer="2081" to-port="2" />
-		<edge from-layer="2080" from-port="0" to-layer="2081" to-port="3" />
-		<edge from-layer="2081" from-port="4" to-layer="2083" to-port="0" />
-		<edge from-layer="2082" from-port="0" to-layer="2083" to-port="2" />
-		<edge from-layer="2083" from-port="3" to-layer="2084" to-port="0" />
-		<edge from-layer="2084" from-port="2" to-layer="2085" to-port="1" />
-		<edge from-layer="2084" from-port="2" to-layer="2126" to-port="1" />
-		<edge from-layer="2085" from-port="2" to-layer="2116" to-port="0" />
-		<edge from-layer="2086" from-port="0" to-layer="2096" to-port="0" />
-		<edge from-layer="2087" from-port="0" to-layer="2096" to-port="1" />
-		<edge from-layer="2088" from-port="1" to-layer="2090" to-port="0" />
-		<edge from-layer="2089" from-port="0" to-layer="2090" to-port="2" />
-		<edge from-layer="2090" from-port="3" to-layer="2091" to-port="0" />
-		<edge from-layer="2091" from-port="2" to-layer="2092" to-port="0" />
-		<edge from-layer="2092" from-port="1" to-layer="2094" to-port="0" />
+		<edge from-layer="2036" from-port="2" to-layer="2038" to-port="0" />
+		<edge from-layer="2037" from-port="0" to-layer="2038" to-port="1" />
+		<edge from-layer="2038" from-port="2" to-layer="2039" to-port="1" />
+		<edge from-layer="2039" from-port="2" to-layer="2040" to-port="0" />
+		<edge from-layer="2040" from-port="2" to-layer="2198" to-port="1" />
+		<edge from-layer="2040" from-port="2" to-layer="2042" to-port="0" />
+		<edge from-layer="2041" from-port="0" to-layer="2042" to-port="1" />
+		<edge from-layer="2042" from-port="2" to-layer="2044" to-port="0" />
+		<edge from-layer="2042" from-port="2" to-layer="2073" to-port="0" />
+		<edge from-layer="2043" from-port="0" to-layer="2044" to-port="1" />
+		<edge from-layer="2044" from-port="2" to-layer="2046" to-port="0" />
+		<edge from-layer="2045" from-port="0" to-layer="2046" to-port="1" />
+		<edge from-layer="2046" from-port="2" to-layer="2055" to-port="0" />
+		<edge from-layer="2047" from-port="0" to-layer="2048" to-port="0" />
+		<edge from-layer="2048" from-port="1" to-layer="2051" to-port="0" />
+		<edge from-layer="2049" from-port="0" to-layer="2050" to-port="0" />
+		<edge from-layer="2050" from-port="1" to-layer="2051" to-port="1" />
+		<edge from-layer="2051" from-port="2" to-layer="2053" to-port="0" />
+		<edge from-layer="2052" from-port="0" to-layer="2053" to-port="1" />
+		<edge from-layer="2053" from-port="2" to-layer="2054" to-port="0" />
+		<edge from-layer="2054" from-port="1" to-layer="2055" to-port="1" />
+		<edge from-layer="2055" from-port="2" to-layer="2058" to-port="0" />
+		<edge from-layer="2056" from-port="0" to-layer="2057" to-port="0" />
+		<edge from-layer="2057" from-port="1" to-layer="2058" to-port="1" />
+		<edge from-layer="2058" from-port="2" to-layer="2059" to-port="0" />
+		<edge from-layer="2059" from-port="1" to-layer="2068" to-port="0" />
+		<edge from-layer="2060" from-port="0" to-layer="2061" to-port="0" />
+		<edge from-layer="2061" from-port="1" to-layer="2064" to-port="0" />
+		<edge from-layer="2062" from-port="0" to-layer="2063" to-port="0" />
+		<edge from-layer="2063" from-port="1" to-layer="2064" to-port="1" />
+		<edge from-layer="2064" from-port="2" to-layer="2066" to-port="0" />
+		<edge from-layer="2065" from-port="0" to-layer="2066" to-port="1" />
+		<edge from-layer="2066" from-port="2" to-layer="2067" to-port="0" />
+		<edge from-layer="2067" from-port="1" to-layer="2068" to-port="1" />
+		<edge from-layer="2068" from-port="2" to-layer="2071" to-port="0" />
+		<edge from-layer="2069" from-port="0" to-layer="2070" to-port="0" />
+		<edge from-layer="2070" from-port="1" to-layer="2071" to-port="1" />
+		<edge from-layer="2071" from-port="2" to-layer="2197" to-port="0" />
+		<edge from-layer="2072" from-port="0" to-layer="2073" to-port="1" />
+		<edge from-layer="2073" from-port="2" to-layer="2075" to-port="0" />
+		<edge from-layer="2074" from-port="0" to-layer="2075" to-port="1" />
+		<edge from-layer="2075" from-port="2" to-layer="2084" to-port="0" />
+		<edge from-layer="2076" from-port="0" to-layer="2077" to-port="0" />
+		<edge from-layer="2077" from-port="1" to-layer="2080" to-port="0" />
+		<edge from-layer="2078" from-port="0" to-layer="2079" to-port="0" />
+		<edge from-layer="2079" from-port="1" to-layer="2080" to-port="1" />
+		<edge from-layer="2080" from-port="2" to-layer="2082" to-port="0" />
+		<edge from-layer="2081" from-port="0" to-layer="2082" to-port="1" />
+		<edge from-layer="2082" from-port="2" to-layer="2083" to-port="0" />
+		<edge from-layer="2083" from-port="1" to-layer="2084" to-port="1" />
+		<edge from-layer="2084" from-port="2" to-layer="2086" to-port="0" />
+		<edge from-layer="2085" from-port="0" to-layer="2086" to-port="1" />
+		<edge from-layer="2086" from-port="2" to-layer="2088" to-port="0" />
+		<edge from-layer="2087" from-port="0" to-layer="2088" to-port="1" />
+		<edge from-layer="2088" from-port="2" to-layer="2092" to-port="0" />
+		<edge from-layer="2088" from-port="2" to-layer="2169" to-port="0" />
+		<edge from-layer="2088" from-port="2" to-layer="2130" to-port="0" />
+		<edge from-layer="2089" from-port="0" to-layer="2092" to-port="1" />
+		<edge from-layer="2090" from-port="0" to-layer="2092" to-port="2" />
+		<edge from-layer="2091" from-port="0" to-layer="2092" to-port="3" />
+		<edge from-layer="2092" from-port="5" to-layer="2094" to-port="0" />
 		<edge from-layer="2093" from-port="0" to-layer="2094" to-port="1" />
-		<edge from-layer="2094" from-port="2" to-layer="2096" to-port="2" />
-		<edge from-layer="2095" from-port="0" to-layer="2096" to-port="3" />
-		<edge from-layer="2096" from-port="4" to-layer="2105" to-port="2" />
-		<edge from-layer="2096" from-port="4" to-layer="2099" to-port="1" />
-		<edge from-layer="2097" from-port="0" to-layer="2099" to-port="2" />
-		<edge from-layer="2098" from-port="0" to-layer="2099" to-port="3" />
-		<edge from-layer="2099" from-port="4" to-layer="2102" to-port="0" />
-		<edge from-layer="2100" from-port="0" to-layer="2101" to-port="0" />
-		<edge from-layer="2101" from-port="1" to-layer="2102" to-port="1" />
-		<edge from-layer="2102" from-port="2" to-layer="2106" to-port="0" />
-		<edge from-layer="2103" from-port="0" to-layer="2105" to-port="1" />
-		<edge from-layer="2104" from-port="0" to-layer="2105" to-port="3" />
-		<edge from-layer="2105" from-port="4" to-layer="2106" to-port="1" />
-		<edge from-layer="2106" from-port="2" to-layer="2115" to-port="0" />
-		<edge from-layer="2107" from-port="0" to-layer="2108" to-port="0" />
-		<edge from-layer="2108" from-port="1" to-layer="2111" to-port="0" />
-		<edge from-layer="2109" from-port="0" to-layer="2111" to-port="1" />
-		<edge from-layer="2110" from-port="0" to-layer="2111" to-port="3" />
-		<edge from-layer="2111" from-port="4" to-layer="2113" to-port="0" />
-		<edge from-layer="2112" from-port="0" to-layer="2113" to-port="2" />
-		<edge from-layer="2113" from-port="3" to-layer="2114" to-port="0" />
-		<edge from-layer="2114" from-port="2" to-layer="2115" to-port="1" />
-		<edge from-layer="2114" from-port="2" to-layer="2148" to-port="1" />
-		<edge from-layer="2115" from-port="2" to-layer="2116" to-port="1" />
-		<edge from-layer="2116" from-port="2" to-layer="2121" to-port="0" />
-		<edge from-layer="2117" from-port="0" to-layer="2120" to-port="1" />
-		<edge from-layer="2118" from-port="0" to-layer="2120" to-port="2" />
-		<edge from-layer="2119" from-port="0" to-layer="2120" to-port="3" />
-		<edge from-layer="2120" from-port="4" to-layer="2121" to-port="1" />
-		<edge from-layer="2121" from-port="2" to-layer="2179" to-port="0" />
-		<edge from-layer="2121" from-port="2" to-layer="2212" to-port="0" />
-		<edge from-layer="2122" from-port="0" to-layer="2125" to-port="1" />
-		<edge from-layer="2123" from-port="0" to-layer="2125" to-port="2" />
-		<edge from-layer="2124" from-port="0" to-layer="2125" to-port="3" />
-		<edge from-layer="2125" from-port="4" to-layer="2140" to-port="0" />
-		<edge from-layer="2125" from-port="4" to-layer="2146" to-port="0" />
-		<edge from-layer="2125" from-port="4" to-layer="2129" to-port="0" />
-		<edge from-layer="2125" from-port="4" to-layer="2126" to-port="0" />
-		<edge from-layer="2126" from-port="2" to-layer="2149" to-port="0" />
-		<edge from-layer="2127" from-port="0" to-layer="2137" to-port="0" />
-		<edge from-layer="2128" from-port="0" to-layer="2137" to-port="1" />
-		<edge from-layer="2129" from-port="1" to-layer="2131" to-port="0" />
-		<edge from-layer="2130" from-port="0" to-layer="2131" to-port="2" />
-		<edge from-layer="2131" from-port="3" to-layer="2132" to-port="0" />
-		<edge from-layer="2132" from-port="2" to-layer="2133" to-port="0" />
-		<edge from-layer="2133" from-port="1" to-layer="2135" to-port="0" />
-		<edge from-layer="2134" from-port="0" to-layer="2135" to-port="1" />
-		<edge from-layer="2135" from-port="2" to-layer="2137" to-port="2" />
-		<edge from-layer="2136" from-port="0" to-layer="2137" to-port="3" />
-		<edge from-layer="2137" from-port="4" to-layer="2140" to-port="1" />
-		<edge from-layer="2137" from-port="4" to-layer="2146" to-port="2" />
-		<edge from-layer="2138" from-port="0" to-layer="2140" to-port="2" />
-		<edge from-layer="2139" from-port="0" to-layer="2140" to-port="3" />
-		<edge from-layer="2140" from-port="4" to-layer="2143" to-port="0" />
-		<edge from-layer="2141" from-port="0" to-layer="2142" to-port="0" />
-		<edge from-layer="2142" from-port="1" to-layer="2143" to-port="1" />
-		<edge from-layer="2143" from-port="2" to-layer="2147" to-port="0" />
-		<edge from-layer="2144" from-port="0" to-layer="2146" to-port="1" />
-		<edge from-layer="2145" from-port="0" to-layer="2146" to-port="3" />
-		<edge from-layer="2146" from-port="4" to-layer="2147" to-port="1" />
-		<edge from-layer="2147" from-port="2" to-layer="2148" to-port="0" />
-		<edge from-layer="2148" from-port="2" to-layer="2149" to-port="1" />
-		<edge from-layer="2149" from-port="2" to-layer="2154" to-port="0" />
-		<edge from-layer="2150" from-port="0" to-layer="2153" to-port="1" />
-		<edge from-layer="2151" from-port="0" to-layer="2153" to-port="2" />
-		<edge from-layer="2152" from-port="0" to-layer="2153" to-port="3" />
-		<edge from-layer="2153" from-port="4" to-layer="2154" to-port="1" />
-		<edge from-layer="2154" from-port="2" to-layer="2155" to-port="1" />
-		<edge from-layer="2155" from-port="2" to-layer="2338" to-port="0" />
-		<edge from-layer="2155" from-port="2" to-layer="2212" to-port="1" />
-		<edge from-layer="2155" from-port="2" to-layer="2176" to-port="0" />
-		<edge from-layer="2156" from-port="0" to-layer="2157" to-port="0" />
-		<edge from-layer="2157" from-port="1" to-layer="2162" to-port="0" />
+		<edge from-layer="2094" from-port="2" to-layer="2116" to-port="0" />
+		<edge from-layer="2094" from-port="2" to-layer="2098" to-port="0" />
+		<edge from-layer="2095" from-port="0" to-layer="2098" to-port="1" />
+		<edge from-layer="2096" from-port="0" to-layer="2098" to-port="2" />
+		<edge from-layer="2097" from-port="0" to-layer="2098" to-port="3" />
+		<edge from-layer="2098" from-port="5" to-layer="2099" to-port="0" />
+		<edge from-layer="2098" from-port="5" to-layer="2103" to-port="0" />
+		<edge from-layer="2098" from-port="5" to-layer="2109" to-port="0" />
+		<edge from-layer="2099" from-port="2" to-layer="2112" to-port="0" />
+		<edge from-layer="2100" from-port="0" to-layer="2103" to-port="1" />
+		<edge from-layer="2101" from-port="0" to-layer="2103" to-port="2" />
+		<edge from-layer="2102" from-port="0" to-layer="2103" to-port="3" />
+		<edge from-layer="2103" from-port="5" to-layer="2105" to-port="0" />
+		<edge from-layer="2104" from-port="0" to-layer="2105" to-port="1" />
+		<edge from-layer="2105" from-port="2" to-layer="2110" to-port="0" />
+		<edge from-layer="2106" from-port="0" to-layer="2109" to-port="1" />
+		<edge from-layer="2107" from-port="0" to-layer="2109" to-port="2" />
+		<edge from-layer="2108" from-port="0" to-layer="2109" to-port="3" />
+		<edge from-layer="2109" from-port="5" to-layer="2110" to-port="1" />
+		<edge from-layer="2110" from-port="2" to-layer="2111" to-port="0" />
+		<edge from-layer="2111" from-port="2" to-layer="2112" to-port="1" />
+		<edge from-layer="2112" from-port="2" to-layer="2117" to-port="0" />
+		<edge from-layer="2113" from-port="0" to-layer="2116" to-port="1" />
+		<edge from-layer="2114" from-port="0" to-layer="2116" to-port="2" />
+		<edge from-layer="2115" from-port="0" to-layer="2116" to-port="3" />
+		<edge from-layer="2116" from-port="5" to-layer="2117" to-port="1" />
+		<edge from-layer="2117" from-port="2" to-layer="2181" to-port="0" />
+		<edge from-layer="2118" from-port="0" to-layer="2123" to-port="0" />
+		<edge from-layer="2119" from-port="0" to-layer="2122" to-port="1" />
+		<edge from-layer="2120" from-port="0" to-layer="2122" to-port="2" />
+		<edge from-layer="2121" from-port="0" to-layer="2122" to-port="3" />
+		<edge from-layer="2122" from-port="4" to-layer="2123" to-port="1" />
+		<edge from-layer="2123" from-port="2" to-layer="2124" to-port="0" />
+		<edge from-layer="2124" from-port="1" to-layer="2126" to-port="0" />
+		<edge from-layer="2125" from-port="0" to-layer="2126" to-port="2" />
+		<edge from-layer="2126" from-port="3" to-layer="2156" to-port="0" />
+		<edge from-layer="2126" from-port="3" to-layer="2174" to-port="0" />
+		<edge from-layer="2127" from-port="0" to-layer="2130" to-port="1" />
+		<edge from-layer="2128" from-port="0" to-layer="2130" to-port="2" />
+		<edge from-layer="2129" from-port="0" to-layer="2130" to-port="3" />
+		<edge from-layer="2130" from-port="5" to-layer="2132" to-port="0" />
+		<edge from-layer="2131" from-port="0" to-layer="2132" to-port="1" />
+		<edge from-layer="2132" from-port="2" to-layer="2154" to-port="0" />
+		<edge from-layer="2132" from-port="2" to-layer="2136" to-port="0" />
+		<edge from-layer="2133" from-port="0" to-layer="2136" to-port="1" />
+		<edge from-layer="2134" from-port="0" to-layer="2136" to-port="2" />
+		<edge from-layer="2135" from-port="0" to-layer="2136" to-port="3" />
+		<edge from-layer="2136" from-port="5" to-layer="2141" to-port="0" />
+		<edge from-layer="2136" from-port="5" to-layer="2137" to-port="0" />
+		<edge from-layer="2136" from-port="5" to-layer="2147" to-port="0" />
+		<edge from-layer="2137" from-port="2" to-layer="2150" to-port="0" />
+		<edge from-layer="2138" from-port="0" to-layer="2141" to-port="1" />
+		<edge from-layer="2139" from-port="0" to-layer="2141" to-port="2" />
+		<edge from-layer="2140" from-port="0" to-layer="2141" to-port="3" />
+		<edge from-layer="2141" from-port="5" to-layer="2143" to-port="0" />
+		<edge from-layer="2142" from-port="0" to-layer="2143" to-port="1" />
+		<edge from-layer="2143" from-port="2" to-layer="2148" to-port="0" />
+		<edge from-layer="2144" from-port="0" to-layer="2147" to-port="1" />
+		<edge from-layer="2145" from-port="0" to-layer="2147" to-port="2" />
+		<edge from-layer="2146" from-port="0" to-layer="2147" to-port="3" />
+		<edge from-layer="2147" from-port="5" to-layer="2148" to-port="1" />
+		<edge from-layer="2148" from-port="2" to-layer="2149" to-port="0" />
+		<edge from-layer="2149" from-port="2" to-layer="2150" to-port="1" />
+		<edge from-layer="2150" from-port="2" to-layer="2155" to-port="0" />
+		<edge from-layer="2151" from-port="0" to-layer="2154" to-port="1" />
+		<edge from-layer="2152" from-port="0" to-layer="2154" to-port="2" />
+		<edge from-layer="2153" from-port="0" to-layer="2154" to-port="3" />
+		<edge from-layer="2154" from-port="5" to-layer="2155" to-port="1" />
+		<edge from-layer="2155" from-port="2" to-layer="2156" to-port="1" />
+		<edge from-layer="2156" from-port="2" to-layer="2181" to-port="1" />
+		<edge from-layer="2156" from-port="2" to-layer="5151" to-port="0" />
+		<edge from-layer="2157" from-port="0" to-layer="2162" to-port="0" />
 		<edge from-layer="2158" from-port="0" to-layer="2161" to-port="1" />
 		<edge from-layer="2159" from-port="0" to-layer="2161" to-port="2" />
 		<edge from-layer="2160" from-port="0" to-layer="2161" to-port="3" />
@@ -123969,212 +83533,215 @@
 		<edge from-layer="2166" from-port="0" to-layer="2169" to-port="1" />
 		<edge from-layer="2167" from-port="0" to-layer="2169" to-port="2" />
 		<edge from-layer="2168" from-port="0" to-layer="2169" to-port="3" />
-		<edge from-layer="2169" from-port="4" to-layer="2171" to-port="0" />
+		<edge from-layer="2169" from-port="5" to-layer="2171" to-port="0" />
 		<edge from-layer="2170" from-port="0" to-layer="2171" to-port="1" />
 		<edge from-layer="2171" from-port="2" to-layer="2172" to-port="1" />
-		<edge from-layer="2172" from-port="2" to-layer="2212" to-port="2" />
-		<edge from-layer="2172" from-port="2" to-layer="2337" to-port="0" />
-		<edge from-layer="2173" from-port="0" to-layer="2191" to-port="0" />
-		<edge from-layer="2174" from-port="0" to-layer="2186" to-port="0" />
-		<edge from-layer="2175" from-port="0" to-layer="2186" to-port="1" />
-		<edge from-layer="2175" from-port="0" to-layer="2189" to-port="1" />
-		<edge from-layer="2176" from-port="1" to-layer="2178" to-port="0" />
-		<edge from-layer="2177" from-port="0" to-layer="2178" to-port="2" />
-		<edge from-layer="2178" from-port="3" to-layer="2182" to-port="0" />
-		<edge from-layer="2178" from-port="3" to-layer="2188" to-port="0" />
-		<edge from-layer="2179" from-port="1" to-layer="2181" to-port="0" />
-		<edge from-layer="2179" from-port="1" to-layer="2207" to-port="0" />
-		<edge from-layer="2180" from-port="0" to-layer="2181" to-port="2" />
-		<edge from-layer="2181" from-port="3" to-layer="2182" to-port="1" />
-		<edge from-layer="2182" from-port="2" to-layer="2184" to-port="0" />
-		<edge from-layer="2183" from-port="0" to-layer="2188" to-port="1" />
-		<edge from-layer="2183" from-port="0" to-layer="2184" to-port="1" />
-		<edge from-layer="2184" from-port="2" to-layer="2186" to-port="2" />
-		<edge from-layer="2185" from-port="0" to-layer="2189" to-port="3" />
-		<edge from-layer="2185" from-port="0" to-layer="2186" to-port="3" />
-		<edge from-layer="2186" from-port="4" to-layer="2191" to-port="1" />
-		<edge from-layer="2187" from-port="0" to-layer="2189" to-port="0" />
-		<edge from-layer="2188" from-port="2" to-layer="2189" to-port="2" />
-		<edge from-layer="2188" from-port="2" to-layer="2196" to-port="2" />
-		<edge from-layer="2189" from-port="4" to-layer="2191" to-port="2" />
-		<edge from-layer="2190" from-port="0" to-layer="2191" to-port="3" />
-		<edge from-layer="2191" from-port="4" to-layer="2198" to-port="0" />
-		<edge from-layer="2192" from-port="0" to-layer="2198" to-port="1" />
-		<edge from-layer="2193" from-port="0" to-layer="2196" to-port="0" />
-		<edge from-layer="2194" from-port="0" to-layer="2196" to-port="1" />
-		<edge from-layer="2195" from-port="0" to-layer="2196" to-port="3" />
-		<edge from-layer="2196" from-port="4" to-layer="2198" to-port="2" />
-		<edge from-layer="2197" from-port="0" to-layer="2198" to-port="3" />
-		<edge from-layer="2198" from-port="4" to-layer="2202" to-port="0" />
-		<edge from-layer="2199" from-port="0" to-layer="2200" to-port="0" />
-		<edge from-layer="2200" from-port="1" to-layer="2202" to-port="1" />
-		<edge from-layer="2201" from-port="0" to-layer="2202" to-port="2" />
-		<edge from-layer="2202" from-port="3" to-layer="2203" to-port="0" />
-		<edge from-layer="2203" from-port="2" to-layer="2204" to-port="0" />
-		<edge from-layer="2203" from-port="2" to-layer="2211" to-port="0" />
-		<edge from-layer="2204" from-port="1" to-layer="2210" to-port="0" />
-		<edge from-layer="2205" from-port="0" to-layer="2207" to-port="1" />
-		<edge from-layer="2206" from-port="0" to-layer="2207" to-port="2" />
-		<edge from-layer="2207" from-port="3" to-layer="2208" to-port="0" />
-		<edge from-layer="2208" from-port="4" to-layer="2209" to-port="0" />
-		<edge from-layer="2209" from-port="1" to-layer="2210" to-port="1" />
-		<edge from-layer="2210" from-port="2" to-layer="2211" to-port="1" />
-		<edge from-layer="2211" from-port="2" to-layer="2212" to-port="3" />
-		<edge from-layer="2212" from-port="4" to-layer="2214" to-port="0" />
-		<edge from-layer="2213" from-port="0" to-layer="2214" to-port="1" />
-		<edge from-layer="2214" from-port="2" to-layer="2216" to-port="0" />
-		<edge from-layer="2215" from-port="0" to-layer="2216" to-port="1" />
-		<edge from-layer="2216" from-port="2" to-layer="2225" to-port="0" />
-		<edge from-layer="2217" from-port="0" to-layer="2218" to-port="0" />
-		<edge from-layer="2218" from-port="1" to-layer="2221" to-port="0" />
-		<edge from-layer="2219" from-port="0" to-layer="2220" to-port="0" />
-		<edge from-layer="2220" from-port="1" to-layer="2221" to-port="1" />
-		<edge from-layer="2221" from-port="2" to-layer="2223" to-port="0" />
-		<edge from-layer="2222" from-port="0" to-layer="2223" to-port="1" />
-		<edge from-layer="2223" from-port="2" to-layer="2224" to-port="0" />
-		<edge from-layer="2224" from-port="1" to-layer="2225" to-port="1" />
-		<edge from-layer="2225" from-port="2" to-layer="2228" to-port="0" />
-		<edge from-layer="2226" from-port="0" to-layer="2227" to-port="0" />
-		<edge from-layer="2227" from-port="1" to-layer="2228" to-port="1" />
-		<edge from-layer="2228" from-port="2" to-layer="2229" to-port="1" />
-		<edge from-layer="2229" from-port="2" to-layer="2230" to-port="0" />
-		<edge from-layer="2230" from-port="2" to-layer="2484" to-port="1" />
-		<edge from-layer="2230" from-port="2" to-layer="2232" to-port="0" />
-		<edge from-layer="2231" from-port="0" to-layer="2232" to-port="1" />
-		<edge from-layer="2232" from-port="2" to-layer="2235" to-port="0" />
-		<edge from-layer="2232" from-port="2" to-layer="2359" to-port="0" />
-		<edge from-layer="2233" from-port="0" to-layer="2234" to-port="0" />
-		<edge from-layer="2234" from-port="1" to-layer="2235" to-port="1" />
-		<edge from-layer="2235" from-port="2" to-layer="2238" to-port="0" />
+		<edge from-layer="2172" from-port="2" to-layer="5150" to-port="0" />
+		<edge from-layer="2172" from-port="2" to-layer="2181" to-port="2" />
+		<edge from-layer="2173" from-port="0" to-layer="2180" to-port="1" />
+		<edge from-layer="2174" from-port="1" to-layer="2177" to-port="0" />
+		<edge from-layer="2175" from-port="0" to-layer="2177" to-port="1" />
+		<edge from-layer="2176" from-port="0" to-layer="2177" to-port="2" />
+		<edge from-layer="2177" from-port="3" to-layer="2178" to-port="0" />
+		<edge from-layer="2178" from-port="2" to-layer="2180" to-port="2" />
+		<edge from-layer="2179" from-port="0" to-layer="2180" to-port="3" />
+		<edge from-layer="2180" from-port="5" to-layer="2181" to-port="3" />
+		<edge from-layer="2181" from-port="4" to-layer="2183" to-port="0" />
+		<edge from-layer="2182" from-port="0" to-layer="2183" to-port="1" />
+		<edge from-layer="2183" from-port="2" to-layer="2185" to-port="0" />
+		<edge from-layer="2184" from-port="0" to-layer="2185" to-port="1" />
+		<edge from-layer="2185" from-port="2" to-layer="2194" to-port="0" />
+		<edge from-layer="2186" from-port="0" to-layer="2187" to-port="0" />
+		<edge from-layer="2187" from-port="1" to-layer="2190" to-port="0" />
+		<edge from-layer="2188" from-port="0" to-layer="2189" to-port="0" />
+		<edge from-layer="2189" from-port="1" to-layer="2190" to-port="1" />
+		<edge from-layer="2190" from-port="2" to-layer="2192" to-port="0" />
+		<edge from-layer="2191" from-port="0" to-layer="2192" to-port="1" />
+		<edge from-layer="2192" from-port="2" to-layer="2193" to-port="0" />
+		<edge from-layer="2193" from-port="1" to-layer="2194" to-port="1" />
+		<edge from-layer="2194" from-port="2" to-layer="2196" to-port="0" />
+		<edge from-layer="2195" from-port="0" to-layer="2196" to-port="1" />
+		<edge from-layer="2196" from-port="2" to-layer="2197" to-port="1" />
+		<edge from-layer="2197" from-port="2" to-layer="2198" to-port="0" />
+		<edge from-layer="2198" from-port="2" to-layer="2356" to-port="1" />
+		<edge from-layer="2198" from-port="2" to-layer="2200" to-port="0" />
+		<edge from-layer="2199" from-port="0" to-layer="2200" to-port="1" />
+		<edge from-layer="2200" from-port="2" to-layer="2202" to-port="0" />
+		<edge from-layer="2200" from-port="2" to-layer="2231" to-port="0" />
+		<edge from-layer="2201" from-port="0" to-layer="2202" to-port="1" />
+		<edge from-layer="2202" from-port="2" to-layer="2204" to-port="0" />
+		<edge from-layer="2203" from-port="0" to-layer="2204" to-port="1" />
+		<edge from-layer="2204" from-port="2" to-layer="2213" to-port="0" />
+		<edge from-layer="2205" from-port="0" to-layer="2206" to-port="0" />
+		<edge from-layer="2206" from-port="1" to-layer="2209" to-port="0" />
+		<edge from-layer="2207" from-port="0" to-layer="2208" to-port="0" />
+		<edge from-layer="2208" from-port="1" to-layer="2209" to-port="1" />
+		<edge from-layer="2209" from-port="2" to-layer="2211" to-port="0" />
+		<edge from-layer="2210" from-port="0" to-layer="2211" to-port="1" />
+		<edge from-layer="2211" from-port="2" to-layer="2212" to-port="0" />
+		<edge from-layer="2212" from-port="1" to-layer="2213" to-port="1" />
+		<edge from-layer="2213" from-port="2" to-layer="2216" to-port="0" />
+		<edge from-layer="2214" from-port="0" to-layer="2215" to-port="0" />
+		<edge from-layer="2215" from-port="1" to-layer="2216" to-port="1" />
+		<edge from-layer="2216" from-port="2" to-layer="2217" to-port="0" />
+		<edge from-layer="2217" from-port="1" to-layer="2226" to-port="0" />
+		<edge from-layer="2218" from-port="0" to-layer="2219" to-port="0" />
+		<edge from-layer="2219" from-port="1" to-layer="2222" to-port="0" />
+		<edge from-layer="2220" from-port="0" to-layer="2221" to-port="0" />
+		<edge from-layer="2221" from-port="1" to-layer="2222" to-port="1" />
+		<edge from-layer="2222" from-port="2" to-layer="2224" to-port="0" />
+		<edge from-layer="2223" from-port="0" to-layer="2224" to-port="1" />
+		<edge from-layer="2224" from-port="2" to-layer="2225" to-port="0" />
+		<edge from-layer="2225" from-port="1" to-layer="2226" to-port="1" />
+		<edge from-layer="2226" from-port="2" to-layer="2229" to-port="0" />
+		<edge from-layer="2227" from-port="0" to-layer="2228" to-port="0" />
+		<edge from-layer="2228" from-port="1" to-layer="2229" to-port="1" />
+		<edge from-layer="2229" from-port="2" to-layer="2355" to-port="0" />
+		<edge from-layer="2230" from-port="0" to-layer="2231" to-port="1" />
+		<edge from-layer="2231" from-port="2" to-layer="2233" to-port="0" />
+		<edge from-layer="2232" from-port="0" to-layer="2233" to-port="1" />
+		<edge from-layer="2233" from-port="2" to-layer="2242" to-port="0" />
+		<edge from-layer="2234" from-port="0" to-layer="2235" to-port="0" />
+		<edge from-layer="2235" from-port="1" to-layer="2238" to-port="0" />
 		<edge from-layer="2236" from-port="0" to-layer="2237" to-port="0" />
 		<edge from-layer="2237" from-port="1" to-layer="2238" to-port="1" />
-		<edge from-layer="2238" from-port="2" to-layer="2247" to-port="0" />
-		<edge from-layer="2239" from-port="0" to-layer="2240" to-port="0" />
-		<edge from-layer="2240" from-port="1" to-layer="2243" to-port="0" />
-		<edge from-layer="2241" from-port="0" to-layer="2242" to-port="0" />
-		<edge from-layer="2242" from-port="1" to-layer="2243" to-port="1" />
-		<edge from-layer="2243" from-port="2" to-layer="2245" to-port="0" />
-		<edge from-layer="2244" from-port="0" to-layer="2245" to-port="1" />
-		<edge from-layer="2245" from-port="2" to-layer="2246" to-port="0" />
-		<edge from-layer="2246" from-port="1" to-layer="2247" to-port="1" />
-		<edge from-layer="2247" from-port="2" to-layer="2250" to-port="0" />
-		<edge from-layer="2248" from-port="0" to-layer="2249" to-port="0" />
-		<edge from-layer="2249" from-port="1" to-layer="2250" to-port="1" />
-		<edge from-layer="2250" from-port="2" to-layer="2252" to-port="0" />
+		<edge from-layer="2238" from-port="2" to-layer="2240" to-port="0" />
+		<edge from-layer="2239" from-port="0" to-layer="2240" to-port="1" />
+		<edge from-layer="2240" from-port="2" to-layer="2241" to-port="0" />
+		<edge from-layer="2241" from-port="1" to-layer="2242" to-port="1" />
+		<edge from-layer="2242" from-port="2" to-layer="2244" to-port="0" />
+		<edge from-layer="2243" from-port="0" to-layer="2244" to-port="1" />
+		<edge from-layer="2244" from-port="2" to-layer="2246" to-port="0" />
+		<edge from-layer="2245" from-port="0" to-layer="2246" to-port="1" />
+		<edge from-layer="2246" from-port="2" to-layer="2327" to-port="0" />
+		<edge from-layer="2246" from-port="2" to-layer="2288" to-port="0" />
+		<edge from-layer="2246" from-port="2" to-layer="2250" to-port="0" />
+		<edge from-layer="2247" from-port="0" to-layer="2250" to-port="1" />
+		<edge from-layer="2248" from-port="0" to-layer="2250" to-port="2" />
+		<edge from-layer="2249" from-port="0" to-layer="2250" to-port="3" />
+		<edge from-layer="2250" from-port="5" to-layer="2252" to-port="0" />
 		<edge from-layer="2251" from-port="0" to-layer="2252" to-port="1" />
-		<edge from-layer="2252" from-port="2" to-layer="2391" to-port="0" />
-		<edge from-layer="2252" from-port="2" to-layer="2256" to-port="0" />
 		<edge from-layer="2252" from-port="2" to-layer="2274" to-port="0" />
+		<edge from-layer="2252" from-port="2" to-layer="2256" to-port="0" />
 		<edge from-layer="2253" from-port="0" to-layer="2256" to-port="1" />
 		<edge from-layer="2254" from-port="0" to-layer="2256" to-port="2" />
 		<edge from-layer="2255" from-port="0" to-layer="2256" to-port="3" />
-		<edge from-layer="2256" from-port="4" to-layer="2258" to-port="0" />
-		<edge from-layer="2257" from-port="0" to-layer="2258" to-port="1" />
-		<edge from-layer="2258" from-port="2" to-layer="2259" to-port="1" />
-		<edge from-layer="2259" from-port="2" to-layer="2260" to-port="0" />
-		<edge from-layer="2259" from-port="2" to-layer="2466" to-port="2" />
-		<edge from-layer="2261" from-port="0" to-layer="2262" to-port="0" />
-		<edge from-layer="2262" from-port="1" to-layer="2267" to-port="0" />
-		<edge from-layer="2263" from-port="0" to-layer="2266" to-port="1" />
-		<edge from-layer="2264" from-port="0" to-layer="2266" to-port="2" />
-		<edge from-layer="2265" from-port="0" to-layer="2266" to-port="3" />
-		<edge from-layer="2266" from-port="4" to-layer="2267" to-port="1" />
-		<edge from-layer="2267" from-port="2" to-layer="2268" to-port="0" />
-		<edge from-layer="2268" from-port="1" to-layer="2270" to-port="0" />
-		<edge from-layer="2269" from-port="0" to-layer="2270" to-port="2" />
-		<edge from-layer="2270" from-port="3" to-layer="2335" to-port="0" />
-		<edge from-layer="2270" from-port="3" to-layer="2287" to-port="0" />
+		<edge from-layer="2256" from-port="5" to-layer="2257" to-port="0" />
+		<edge from-layer="2256" from-port="5" to-layer="2267" to-port="0" />
+		<edge from-layer="2256" from-port="5" to-layer="2261" to-port="0" />
+		<edge from-layer="2257" from-port="2" to-layer="2270" to-port="0" />
+		<edge from-layer="2258" from-port="0" to-layer="2261" to-port="1" />
+		<edge from-layer="2259" from-port="0" to-layer="2261" to-port="2" />
+		<edge from-layer="2260" from-port="0" to-layer="2261" to-port="3" />
+		<edge from-layer="2261" from-port="5" to-layer="2263" to-port="0" />
+		<edge from-layer="2262" from-port="0" to-layer="2263" to-port="1" />
+		<edge from-layer="2263" from-port="2" to-layer="2268" to-port="0" />
+		<edge from-layer="2264" from-port="0" to-layer="2267" to-port="1" />
+		<edge from-layer="2265" from-port="0" to-layer="2267" to-port="2" />
+		<edge from-layer="2266" from-port="0" to-layer="2267" to-port="3" />
+		<edge from-layer="2267" from-port="5" to-layer="2268" to-port="1" />
+		<edge from-layer="2268" from-port="2" to-layer="2269" to-port="0" />
+		<edge from-layer="2269" from-port="2" to-layer="2270" to-port="1" />
+		<edge from-layer="2270" from-port="2" to-layer="2275" to-port="0" />
 		<edge from-layer="2271" from-port="0" to-layer="2274" to-port="1" />
 		<edge from-layer="2272" from-port="0" to-layer="2274" to-port="2" />
 		<edge from-layer="2273" from-port="0" to-layer="2274" to-port="3" />
-		<edge from-layer="2274" from-port="4" to-layer="2276" to-port="0" />
-		<edge from-layer="2275" from-port="0" to-layer="2276" to-port="1" />
-		<edge from-layer="2276" from-port="2" to-layer="2333" to-port="0" />
-		<edge from-layer="2276" from-port="2" to-layer="2284" to-port="0" />
-		<edge from-layer="2276" from-port="2" to-layer="2280" to-port="0" />
+		<edge from-layer="2274" from-port="5" to-layer="2275" to-port="1" />
+		<edge from-layer="2275" from-port="2" to-layer="2339" to-port="0" />
+		<edge from-layer="2276" from-port="0" to-layer="2281" to-port="0" />
 		<edge from-layer="2277" from-port="0" to-layer="2280" to-port="1" />
 		<edge from-layer="2278" from-port="0" to-layer="2280" to-port="2" />
 		<edge from-layer="2279" from-port="0" to-layer="2280" to-port="3" />
-		<edge from-layer="2280" from-port="4" to-layer="2312" to-port="0" />
-		<edge from-layer="2280" from-port="4" to-layer="2298" to-port="0" />
-		<edge from-layer="2280" from-port="4" to-layer="2301" to-port="0" />
-		<edge from-layer="2280" from-port="4" to-layer="2318" to-port="0" />
-		<edge from-layer="2281" from-port="0" to-layer="2282" to-port="0" />
-		<edge from-layer="2282" from-port="1" to-layer="2294" to-port="0" />
-		<edge from-layer="2283" from-port="0" to-layer="2294" to-port="1" />
-		<edge from-layer="2284" from-port="1" to-layer="2286" to-port="0" />
-		<edge from-layer="2285" from-port="0" to-layer="2286" to-port="2" />
-		<edge from-layer="2286" from-port="3" to-layer="2290" to-port="0" />
-		<edge from-layer="2287" from-port="1" to-layer="2289" to-port="0" />
-		<edge from-layer="2288" from-port="0" to-layer="2289" to-port="2" />
-		<edge from-layer="2289" from-port="3" to-layer="2290" to-port="1" />
-		<edge from-layer="2290" from-port="2" to-layer="2292" to-port="0" />
-		<edge from-layer="2291" from-port="0" to-layer="2292" to-port="1" />
-		<edge from-layer="2292" from-port="2" to-layer="2294" to-port="2" />
-		<edge from-layer="2292" from-port="2" to-layer="2324" to-port="2" />
+		<edge from-layer="2280" from-port="4" to-layer="2281" to-port="1" />
+		<edge from-layer="2281" from-port="2" to-layer="2282" to-port="0" />
+		<edge from-layer="2282" from-port="1" to-layer="2284" to-port="0" />
+		<edge from-layer="2283" from-port="0" to-layer="2284" to-port="2" />
+		<edge from-layer="2284" from-port="3" to-layer="2314" to-port="0" />
+		<edge from-layer="2284" from-port="3" to-layer="2332" to-port="0" />
+		<edge from-layer="2285" from-port="0" to-layer="2288" to-port="1" />
+		<edge from-layer="2286" from-port="0" to-layer="2288" to-port="2" />
+		<edge from-layer="2287" from-port="0" to-layer="2288" to-port="3" />
+		<edge from-layer="2288" from-port="5" to-layer="2290" to-port="0" />
+		<edge from-layer="2289" from-port="0" to-layer="2290" to-port="1" />
+		<edge from-layer="2290" from-port="2" to-layer="2294" to-port="0" />
+		<edge from-layer="2290" from-port="2" to-layer="2312" to-port="0" />
+		<edge from-layer="2291" from-port="0" to-layer="2294" to-port="1" />
+		<edge from-layer="2292" from-port="0" to-layer="2294" to-port="2" />
 		<edge from-layer="2293" from-port="0" to-layer="2294" to-port="3" />
-		<edge from-layer="2294" from-port="4" to-layer="2296" to-port="0" />
-		<edge from-layer="2295" from-port="0" to-layer="2296" to-port="2" />
-		<edge from-layer="2296" from-port="3" to-layer="2297" to-port="0" />
-		<edge from-layer="2297" from-port="2" to-layer="2398" to-port="1" />
-		<edge from-layer="2297" from-port="2" to-layer="2298" to-port="1" />
-		<edge from-layer="2298" from-port="2" to-layer="2329" to-port="0" />
-		<edge from-layer="2299" from-port="0" to-layer="2309" to-port="0" />
-		<edge from-layer="2300" from-port="0" to-layer="2309" to-port="1" />
-		<edge from-layer="2301" from-port="1" to-layer="2303" to-port="0" />
-		<edge from-layer="2302" from-port="0" to-layer="2303" to-port="2" />
-		<edge from-layer="2303" from-port="3" to-layer="2304" to-port="0" />
-		<edge from-layer="2304" from-port="2" to-layer="2305" to-port="0" />
-		<edge from-layer="2305" from-port="1" to-layer="2307" to-port="0" />
-		<edge from-layer="2306" from-port="0" to-layer="2307" to-port="1" />
-		<edge from-layer="2307" from-port="2" to-layer="2309" to-port="2" />
-		<edge from-layer="2308" from-port="0" to-layer="2309" to-port="3" />
-		<edge from-layer="2309" from-port="4" to-layer="2318" to-port="2" />
-		<edge from-layer="2309" from-port="4" to-layer="2312" to-port="1" />
+		<edge from-layer="2294" from-port="5" to-layer="2305" to-port="0" />
+		<edge from-layer="2294" from-port="5" to-layer="2295" to-port="0" />
+		<edge from-layer="2294" from-port="5" to-layer="2299" to-port="0" />
+		<edge from-layer="2295" from-port="2" to-layer="2308" to-port="0" />
+		<edge from-layer="2296" from-port="0" to-layer="2299" to-port="1" />
+		<edge from-layer="2297" from-port="0" to-layer="2299" to-port="2" />
+		<edge from-layer="2298" from-port="0" to-layer="2299" to-port="3" />
+		<edge from-layer="2299" from-port="5" to-layer="2301" to-port="0" />
+		<edge from-layer="2300" from-port="0" to-layer="2301" to-port="1" />
+		<edge from-layer="2301" from-port="2" to-layer="2306" to-port="0" />
+		<edge from-layer="2302" from-port="0" to-layer="2305" to-port="1" />
+		<edge from-layer="2303" from-port="0" to-layer="2305" to-port="2" />
+		<edge from-layer="2304" from-port="0" to-layer="2305" to-port="3" />
+		<edge from-layer="2305" from-port="5" to-layer="2306" to-port="1" />
+		<edge from-layer="2306" from-port="2" to-layer="2307" to-port="0" />
+		<edge from-layer="2307" from-port="2" to-layer="2308" to-port="1" />
+		<edge from-layer="2308" from-port="2" to-layer="2313" to-port="0" />
+		<edge from-layer="2309" from-port="0" to-layer="2312" to-port="1" />
 		<edge from-layer="2310" from-port="0" to-layer="2312" to-port="2" />
 		<edge from-layer="2311" from-port="0" to-layer="2312" to-port="3" />
-		<edge from-layer="2312" from-port="4" to-layer="2315" to-port="0" />
-		<edge from-layer="2313" from-port="0" to-layer="2314" to-port="0" />
-		<edge from-layer="2314" from-port="1" to-layer="2315" to-port="1" />
-		<edge from-layer="2315" from-port="2" to-layer="2319" to-port="0" />
-		<edge from-layer="2316" from-port="0" to-layer="2318" to-port="1" />
-		<edge from-layer="2317" from-port="0" to-layer="2318" to-port="3" />
-		<edge from-layer="2318" from-port="4" to-layer="2319" to-port="1" />
-		<edge from-layer="2319" from-port="2" to-layer="2328" to-port="0" />
-		<edge from-layer="2320" from-port="0" to-layer="2321" to-port="0" />
-		<edge from-layer="2321" from-port="1" to-layer="2324" to-port="0" />
-		<edge from-layer="2322" from-port="0" to-layer="2324" to-port="1" />
-		<edge from-layer="2323" from-port="0" to-layer="2324" to-port="3" />
-		<edge from-layer="2324" from-port="4" to-layer="2326" to-port="0" />
-		<edge from-layer="2325" from-port="0" to-layer="2326" to-port="2" />
-		<edge from-layer="2326" from-port="3" to-layer="2327" to-port="0" />
-		<edge from-layer="2327" from-port="2" to-layer="2420" to-port="1" />
-		<edge from-layer="2327" from-port="2" to-layer="2328" to-port="1" />
-		<edge from-layer="2328" from-port="2" to-layer="2329" to-port="1" />
-		<edge from-layer="2329" from-port="2" to-layer="2334" to-port="0" />
-		<edge from-layer="2330" from-port="0" to-layer="2333" to-port="1" />
-		<edge from-layer="2331" from-port="0" to-layer="2333" to-port="2" />
-		<edge from-layer="2332" from-port="0" to-layer="2333" to-port="3" />
-		<edge from-layer="2333" from-port="4" to-layer="2334" to-port="1" />
-		<edge from-layer="2334" from-port="2" to-layer="2335" to-port="1" />
-		<edge from-layer="2335" from-port="2" to-layer="2466" to-port="1" />
-		<edge from-layer="2335" from-port="2" to-layer="2430" to-port="0" />
-		<edge from-layer="2335" from-port="2" to-layer="2336" to-port="0" />
-		<edge from-layer="2347" from-port="0" to-layer="2348" to-port="0" />
-		<edge from-layer="2348" from-port="1" to-layer="2353" to-port="0" />
-		<edge from-layer="2349" from-port="0" to-layer="2352" to-port="1" />
-		<edge from-layer="2350" from-port="0" to-layer="2352" to-port="2" />
-		<edge from-layer="2351" from-port="0" to-layer="2352" to-port="3" />
-		<edge from-layer="2352" from-port="4" to-layer="2353" to-port="1" />
-		<edge from-layer="2353" from-port="2" to-layer="2354" to-port="0" />
-		<edge from-layer="2354" from-port="1" to-layer="2356" to-port="0" />
-		<edge from-layer="2355" from-port="0" to-layer="2356" to-port="2" />
-		<edge from-layer="2356" from-port="3" to-layer="7595" to-port="0" />
-		<edge from-layer="2357" from-port="0" to-layer="2358" to-port="0" />
-		<edge from-layer="2358" from-port="1" to-layer="2359" to-port="1" />
-		<edge from-layer="2359" from-port="2" to-layer="2362" to-port="0" />
-		<edge from-layer="2360" from-port="0" to-layer="2361" to-port="0" />
-		<edge from-layer="2361" from-port="1" to-layer="2362" to-port="1" />
+		<edge from-layer="2312" from-port="5" to-layer="2313" to-port="1" />
+		<edge from-layer="2313" from-port="2" to-layer="2314" to-port="1" />
+		<edge from-layer="2314" from-port="2" to-layer="5149" to-port="0" />
+		<edge from-layer="2314" from-port="2" to-layer="2339" to-port="1" />
+		<edge from-layer="2315" from-port="0" to-layer="2320" to-port="0" />
+		<edge from-layer="2316" from-port="0" to-layer="2319" to-port="1" />
+		<edge from-layer="2317" from-port="0" to-layer="2319" to-port="2" />
+		<edge from-layer="2318" from-port="0" to-layer="2319" to-port="3" />
+		<edge from-layer="2319" from-port="4" to-layer="2320" to-port="1" />
+		<edge from-layer="2320" from-port="2" to-layer="2321" to-port="0" />
+		<edge from-layer="2321" from-port="1" to-layer="2323" to-port="0" />
+		<edge from-layer="2322" from-port="0" to-layer="2323" to-port="2" />
+		<edge from-layer="2323" from-port="3" to-layer="2330" to-port="0" />
+		<edge from-layer="2324" from-port="0" to-layer="2327" to-port="1" />
+		<edge from-layer="2325" from-port="0" to-layer="2327" to-port="2" />
+		<edge from-layer="2326" from-port="0" to-layer="2327" to-port="3" />
+		<edge from-layer="2327" from-port="5" to-layer="2329" to-port="0" />
+		<edge from-layer="2328" from-port="0" to-layer="2329" to-port="1" />
+		<edge from-layer="2329" from-port="2" to-layer="2330" to-port="1" />
+		<edge from-layer="2330" from-port="2" to-layer="5148" to-port="0" />
+		<edge from-layer="2330" from-port="2" to-layer="2339" to-port="2" />
+		<edge from-layer="2331" from-port="0" to-layer="2338" to-port="1" />
+		<edge from-layer="2332" from-port="1" to-layer="2335" to-port="0" />
+		<edge from-layer="2333" from-port="0" to-layer="2335" to-port="1" />
+		<edge from-layer="2334" from-port="0" to-layer="2335" to-port="2" />
+		<edge from-layer="2335" from-port="3" to-layer="2336" to-port="0" />
+		<edge from-layer="2336" from-port="2" to-layer="2338" to-port="2" />
+		<edge from-layer="2337" from-port="0" to-layer="2338" to-port="3" />
+		<edge from-layer="2338" from-port="5" to-layer="2339" to-port="3" />
+		<edge from-layer="2339" from-port="4" to-layer="2341" to-port="0" />
+		<edge from-layer="2340" from-port="0" to-layer="2341" to-port="1" />
+		<edge from-layer="2341" from-port="2" to-layer="2343" to-port="0" />
+		<edge from-layer="2342" from-port="0" to-layer="2343" to-port="1" />
+		<edge from-layer="2343" from-port="2" to-layer="2352" to-port="0" />
+		<edge from-layer="2344" from-port="0" to-layer="2345" to-port="0" />
+		<edge from-layer="2345" from-port="1" to-layer="2348" to-port="0" />
+		<edge from-layer="2346" from-port="0" to-layer="2347" to-port="0" />
+		<edge from-layer="2347" from-port="1" to-layer="2348" to-port="1" />
+		<edge from-layer="2348" from-port="2" to-layer="2350" to-port="0" />
+		<edge from-layer="2349" from-port="0" to-layer="2350" to-port="1" />
+		<edge from-layer="2350" from-port="2" to-layer="2351" to-port="0" />
+		<edge from-layer="2351" from-port="1" to-layer="2352" to-port="1" />
+		<edge from-layer="2352" from-port="2" to-layer="2354" to-port="0" />
+		<edge from-layer="2353" from-port="0" to-layer="2354" to-port="1" />
+		<edge from-layer="2354" from-port="2" to-layer="2355" to-port="1" />
+		<edge from-layer="2355" from-port="2" to-layer="2356" to-port="0" />
+		<edge from-layer="2356" from-port="2" to-layer="2514" to-port="1" />
+		<edge from-layer="2356" from-port="2" to-layer="2358" to-port="0" />
+		<edge from-layer="2357" from-port="0" to-layer="2358" to-port="1" />
+		<edge from-layer="2358" from-port="2" to-layer="2360" to-port="0" />
+		<edge from-layer="2358" from-port="2" to-layer="2389" to-port="0" />
+		<edge from-layer="2359" from-port="0" to-layer="2360" to-port="1" />
+		<edge from-layer="2360" from-port="2" to-layer="2362" to-port="0" />
+		<edge from-layer="2361" from-port="0" to-layer="2362" to-port="1" />
 		<edge from-layer="2362" from-port="2" to-layer="2371" to-port="0" />
 		<edge from-layer="2363" from-port="0" to-layer="2364" to-port="0" />
 		<edge from-layer="2364" from-port="1" to-layer="2367" to-port="0" />
@@ -124200,702 +83767,679 @@
 		<edge from-layer="2384" from-port="2" to-layer="2387" to-port="0" />
 		<edge from-layer="2385" from-port="0" to-layer="2386" to-port="0" />
 		<edge from-layer="2386" from-port="1" to-layer="2387" to-port="1" />
-		<edge from-layer="2387" from-port="2" to-layer="2483" to-port="0" />
-		<edge from-layer="2388" from-port="0" to-layer="2391" to-port="1" />
-		<edge from-layer="2389" from-port="0" to-layer="2391" to-port="2" />
-		<edge from-layer="2390" from-port="0" to-layer="2391" to-port="3" />
-		<edge from-layer="2391" from-port="4" to-layer="2393" to-port="0" />
-		<edge from-layer="2392" from-port="0" to-layer="2393" to-port="1" />
-		<edge from-layer="2393" from-port="2" to-layer="2397" to-port="0" />
-		<edge from-layer="2393" from-port="2" to-layer="2425" to-port="0" />
-		<edge from-layer="2394" from-port="0" to-layer="2397" to-port="1" />
-		<edge from-layer="2395" from-port="0" to-layer="2397" to-port="2" />
-		<edge from-layer="2396" from-port="0" to-layer="2397" to-port="3" />
-		<edge from-layer="2397" from-port="4" to-layer="2401" to-port="0" />
-		<edge from-layer="2397" from-port="4" to-layer="2418" to-port="0" />
-		<edge from-layer="2397" from-port="4" to-layer="2412" to-port="0" />
-		<edge from-layer="2397" from-port="4" to-layer="2398" to-port="0" />
-		<edge from-layer="2398" from-port="2" to-layer="2421" to-port="0" />
-		<edge from-layer="2399" from-port="0" to-layer="2409" to-port="0" />
-		<edge from-layer="2400" from-port="0" to-layer="2409" to-port="1" />
-		<edge from-layer="2401" from-port="1" to-layer="2403" to-port="0" />
-		<edge from-layer="2402" from-port="0" to-layer="2403" to-port="2" />
-		<edge from-layer="2403" from-port="3" to-layer="2404" to-port="0" />
-		<edge from-layer="2404" from-port="2" to-layer="2405" to-port="0" />
-		<edge from-layer="2405" from-port="1" to-layer="2407" to-port="0" />
-		<edge from-layer="2406" from-port="0" to-layer="2407" to-port="1" />
-		<edge from-layer="2407" from-port="2" to-layer="2409" to-port="2" />
-		<edge from-layer="2408" from-port="0" to-layer="2409" to-port="3" />
-		<edge from-layer="2409" from-port="4" to-layer="2418" to-port="2" />
-		<edge from-layer="2409" from-port="4" to-layer="2412" to-port="1" />
-		<edge from-layer="2410" from-port="0" to-layer="2412" to-port="2" />
-		<edge from-layer="2411" from-port="0" to-layer="2412" to-port="3" />
-		<edge from-layer="2412" from-port="4" to-layer="2415" to-port="0" />
-		<edge from-layer="2413" from-port="0" to-layer="2414" to-port="0" />
-		<edge from-layer="2414" from-port="1" to-layer="2415" to-port="1" />
-		<edge from-layer="2415" from-port="2" to-layer="2419" to-port="0" />
-		<edge from-layer="2416" from-port="0" to-layer="2418" to-port="1" />
-		<edge from-layer="2417" from-port="0" to-layer="2418" to-port="3" />
-		<edge from-layer="2418" from-port="4" to-layer="2419" to-port="1" />
-		<edge from-layer="2419" from-port="2" to-layer="2420" to-port="0" />
-		<edge from-layer="2420" from-port="2" to-layer="2421" to-port="1" />
+		<edge from-layer="2387" from-port="2" to-layer="2513" to-port="0" />
+		<edge from-layer="2388" from-port="0" to-layer="2389" to-port="1" />
+		<edge from-layer="2389" from-port="2" to-layer="2391" to-port="0" />
+		<edge from-layer="2390" from-port="0" to-layer="2391" to-port="1" />
+		<edge from-layer="2391" from-port="2" to-layer="2400" to-port="0" />
+		<edge from-layer="2392" from-port="0" to-layer="2393" to-port="0" />
+		<edge from-layer="2393" from-port="1" to-layer="2396" to-port="0" />
+		<edge from-layer="2394" from-port="0" to-layer="2395" to-port="0" />
+		<edge from-layer="2395" from-port="1" to-layer="2396" to-port="1" />
+		<edge from-layer="2396" from-port="2" to-layer="2398" to-port="0" />
+		<edge from-layer="2397" from-port="0" to-layer="2398" to-port="1" />
+		<edge from-layer="2398" from-port="2" to-layer="2399" to-port="0" />
+		<edge from-layer="2399" from-port="1" to-layer="2400" to-port="1" />
+		<edge from-layer="2400" from-port="2" to-layer="2402" to-port="0" />
+		<edge from-layer="2401" from-port="0" to-layer="2402" to-port="1" />
+		<edge from-layer="2402" from-port="2" to-layer="2404" to-port="0" />
+		<edge from-layer="2403" from-port="0" to-layer="2404" to-port="1" />
+		<edge from-layer="2404" from-port="2" to-layer="2446" to-port="0" />
+		<edge from-layer="2404" from-port="2" to-layer="2408" to-port="0" />
+		<edge from-layer="2404" from-port="2" to-layer="2485" to-port="0" />
+		<edge from-layer="2405" from-port="0" to-layer="2408" to-port="1" />
+		<edge from-layer="2406" from-port="0" to-layer="2408" to-port="2" />
+		<edge from-layer="2407" from-port="0" to-layer="2408" to-port="3" />
+		<edge from-layer="2408" from-port="5" to-layer="2410" to-port="0" />
+		<edge from-layer="2409" from-port="0" to-layer="2410" to-port="1" />
+		<edge from-layer="2410" from-port="2" to-layer="2414" to-port="0" />
+		<edge from-layer="2410" from-port="2" to-layer="2432" to-port="0" />
+		<edge from-layer="2411" from-port="0" to-layer="2414" to-port="1" />
+		<edge from-layer="2412" from-port="0" to-layer="2414" to-port="2" />
+		<edge from-layer="2413" from-port="0" to-layer="2414" to-port="3" />
+		<edge from-layer="2414" from-port="5" to-layer="2425" to-port="0" />
+		<edge from-layer="2414" from-port="5" to-layer="2415" to-port="0" />
+		<edge from-layer="2414" from-port="5" to-layer="2419" to-port="0" />
+		<edge from-layer="2415" from-port="2" to-layer="2428" to-port="0" />
+		<edge from-layer="2416" from-port="0" to-layer="2419" to-port="1" />
+		<edge from-layer="2417" from-port="0" to-layer="2419" to-port="2" />
+		<edge from-layer="2418" from-port="0" to-layer="2419" to-port="3" />
+		<edge from-layer="2419" from-port="5" to-layer="2421" to-port="0" />
+		<edge from-layer="2420" from-port="0" to-layer="2421" to-port="1" />
 		<edge from-layer="2421" from-port="2" to-layer="2426" to-port="0" />
 		<edge from-layer="2422" from-port="0" to-layer="2425" to-port="1" />
 		<edge from-layer="2423" from-port="0" to-layer="2425" to-port="2" />
 		<edge from-layer="2424" from-port="0" to-layer="2425" to-port="3" />
-		<edge from-layer="2425" from-port="4" to-layer="2426" to-port="1" />
-		<edge from-layer="2426" from-port="2" to-layer="2433" to-port="0" />
-		<edge from-layer="2426" from-port="2" to-layer="2466" to-port="0" />
-		<edge from-layer="2427" from-port="0" to-layer="2445" to-port="0" />
-		<edge from-layer="2428" from-port="0" to-layer="2440" to-port="0" />
-		<edge from-layer="2429" from-port="0" to-layer="2443" to-port="1" />
-		<edge from-layer="2429" from-port="0" to-layer="2440" to-port="1" />
-		<edge from-layer="2430" from-port="1" to-layer="2432" to-port="0" />
-		<edge from-layer="2431" from-port="0" to-layer="2432" to-port="2" />
-		<edge from-layer="2432" from-port="3" to-layer="2442" to-port="0" />
-		<edge from-layer="2432" from-port="3" to-layer="2436" to-port="0" />
-		<edge from-layer="2433" from-port="1" to-layer="2435" to-port="0" />
-		<edge from-layer="2433" from-port="1" to-layer="2461" to-port="0" />
-		<edge from-layer="2434" from-port="0" to-layer="2435" to-port="2" />
-		<edge from-layer="2435" from-port="3" to-layer="2436" to-port="1" />
-		<edge from-layer="2436" from-port="2" to-layer="2438" to-port="0" />
-		<edge from-layer="2437" from-port="0" to-layer="2438" to-port="1" />
-		<edge from-layer="2437" from-port="0" to-layer="2442" to-port="1" />
-		<edge from-layer="2438" from-port="2" to-layer="2440" to-port="2" />
-		<edge from-layer="2439" from-port="0" to-layer="2443" to-port="3" />
-		<edge from-layer="2439" from-port="0" to-layer="2440" to-port="3" />
-		<edge from-layer="2440" from-port="4" to-layer="2445" to-port="1" />
-		<edge from-layer="2441" from-port="0" to-layer="2443" to-port="0" />
-		<edge from-layer="2442" from-port="2" to-layer="2450" to-port="2" />
-		<edge from-layer="2442" from-port="2" to-layer="2443" to-port="2" />
-		<edge from-layer="2443" from-port="4" to-layer="2445" to-port="2" />
-		<edge from-layer="2444" from-port="0" to-layer="2445" to-port="3" />
-		<edge from-layer="2445" from-port="4" to-layer="2452" to-port="0" />
-		<edge from-layer="2446" from-port="0" to-layer="2452" to-port="1" />
-		<edge from-layer="2447" from-port="0" to-layer="2450" to-port="0" />
-		<edge from-layer="2448" from-port="0" to-layer="2450" to-port="1" />
-		<edge from-layer="2449" from-port="0" to-layer="2450" to-port="3" />
-		<edge from-layer="2450" from-port="4" to-layer="2452" to-port="2" />
+		<edge from-layer="2425" from-port="5" to-layer="2426" to-port="1" />
+		<edge from-layer="2426" from-port="2" to-layer="2427" to-port="0" />
+		<edge from-layer="2427" from-port="2" to-layer="2428" to-port="1" />
+		<edge from-layer="2428" from-port="2" to-layer="2433" to-port="0" />
+		<edge from-layer="2429" from-port="0" to-layer="2432" to-port="1" />
+		<edge from-layer="2430" from-port="0" to-layer="2432" to-port="2" />
+		<edge from-layer="2431" from-port="0" to-layer="2432" to-port="3" />
+		<edge from-layer="2432" from-port="5" to-layer="2433" to-port="1" />
+		<edge from-layer="2433" from-port="2" to-layer="2497" to-port="0" />
+		<edge from-layer="2434" from-port="0" to-layer="2439" to-port="0" />
+		<edge from-layer="2435" from-port="0" to-layer="2438" to-port="1" />
+		<edge from-layer="2436" from-port="0" to-layer="2438" to-port="2" />
+		<edge from-layer="2437" from-port="0" to-layer="2438" to-port="3" />
+		<edge from-layer="2438" from-port="4" to-layer="2439" to-port="1" />
+		<edge from-layer="2439" from-port="2" to-layer="2440" to-port="0" />
+		<edge from-layer="2440" from-port="1" to-layer="2442" to-port="0" />
+		<edge from-layer="2441" from-port="0" to-layer="2442" to-port="2" />
+		<edge from-layer="2442" from-port="3" to-layer="2472" to-port="0" />
+		<edge from-layer="2442" from-port="3" to-layer="2490" to-port="0" />
+		<edge from-layer="2443" from-port="0" to-layer="2446" to-port="1" />
+		<edge from-layer="2444" from-port="0" to-layer="2446" to-port="2" />
+		<edge from-layer="2445" from-port="0" to-layer="2446" to-port="3" />
+		<edge from-layer="2446" from-port="5" to-layer="2448" to-port="0" />
+		<edge from-layer="2447" from-port="0" to-layer="2448" to-port="1" />
+		<edge from-layer="2448" from-port="2" to-layer="2470" to-port="0" />
+		<edge from-layer="2448" from-port="2" to-layer="2452" to-port="0" />
+		<edge from-layer="2449" from-port="0" to-layer="2452" to-port="1" />
+		<edge from-layer="2450" from-port="0" to-layer="2452" to-port="2" />
 		<edge from-layer="2451" from-port="0" to-layer="2452" to-port="3" />
-		<edge from-layer="2452" from-port="4" to-layer="2456" to-port="0" />
-		<edge from-layer="2453" from-port="0" to-layer="2454" to-port="0" />
-		<edge from-layer="2454" from-port="1" to-layer="2456" to-port="1" />
-		<edge from-layer="2455" from-port="0" to-layer="2456" to-port="2" />
-		<edge from-layer="2456" from-port="3" to-layer="2457" to-port="0" />
-		<edge from-layer="2457" from-port="2" to-layer="2458" to-port="0" />
-		<edge from-layer="2457" from-port="2" to-layer="2465" to-port="0" />
-		<edge from-layer="2458" from-port="1" to-layer="2464" to-port="0" />
-		<edge from-layer="2459" from-port="0" to-layer="2461" to-port="1" />
-		<edge from-layer="2460" from-port="0" to-layer="2461" to-port="2" />
-		<edge from-layer="2461" from-port="3" to-layer="2462" to-port="0" />
-		<edge from-layer="2462" from-port="4" to-layer="2463" to-port="0" />
-		<edge from-layer="2463" from-port="1" to-layer="2464" to-port="1" />
-		<edge from-layer="2464" from-port="2" to-layer="2465" to-port="1" />
-		<edge from-layer="2465" from-port="2" to-layer="2466" to-port="3" />
-		<edge from-layer="2466" from-port="4" to-layer="2468" to-port="0" />
-		<edge from-layer="2467" from-port="0" to-layer="2468" to-port="1" />
-		<edge from-layer="2468" from-port="2" to-layer="2470" to-port="0" />
-		<edge from-layer="2469" from-port="0" to-layer="2470" to-port="1" />
-		<edge from-layer="2470" from-port="2" to-layer="2479" to-port="0" />
-		<edge from-layer="2471" from-port="0" to-layer="2472" to-port="0" />
-		<edge from-layer="2472" from-port="1" to-layer="2475" to-port="0" />
-		<edge from-layer="2473" from-port="0" to-layer="2474" to-port="0" />
-		<edge from-layer="2474" from-port="1" to-layer="2475" to-port="1" />
-		<edge from-layer="2475" from-port="2" to-layer="2477" to-port="0" />
-		<edge from-layer="2476" from-port="0" to-layer="2477" to-port="1" />
-		<edge from-layer="2477" from-port="2" to-layer="2478" to-port="0" />
-		<edge from-layer="2478" from-port="1" to-layer="2479" to-port="1" />
-		<edge from-layer="2479" from-port="2" to-layer="2482" to-port="0" />
-		<edge from-layer="2480" from-port="0" to-layer="2481" to-port="0" />
-		<edge from-layer="2481" from-port="1" to-layer="2482" to-port="1" />
-		<edge from-layer="2482" from-port="2" to-layer="2483" to-port="1" />
-		<edge from-layer="2483" from-port="2" to-layer="2484" to-port="0" />
-		<edge from-layer="2484" from-port="2" to-layer="2486" to-port="0" />
-		<edge from-layer="2484" from-port="2" to-layer="2726" to-port="1" />
-		<edge from-layer="2485" from-port="0" to-layer="2486" to-port="1" />
-		<edge from-layer="2486" from-port="2" to-layer="2489" to-port="0" />
-		<edge from-layer="2486" from-port="2" to-layer="2520" to-port="0" />
-		<edge from-layer="2487" from-port="0" to-layer="2488" to-port="0" />
-		<edge from-layer="2488" from-port="1" to-layer="2489" to-port="1" />
-		<edge from-layer="2489" from-port="2" to-layer="2492" to-port="0" />
-		<edge from-layer="2490" from-port="0" to-layer="2491" to-port="0" />
-		<edge from-layer="2491" from-port="1" to-layer="2492" to-port="1" />
-		<edge from-layer="2492" from-port="2" to-layer="2501" to-port="0" />
-		<edge from-layer="2493" from-port="0" to-layer="2494" to-port="0" />
-		<edge from-layer="2494" from-port="1" to-layer="2497" to-port="0" />
-		<edge from-layer="2495" from-port="0" to-layer="2496" to-port="0" />
-		<edge from-layer="2496" from-port="1" to-layer="2497" to-port="1" />
-		<edge from-layer="2497" from-port="2" to-layer="2499" to-port="0" />
+		<edge from-layer="2452" from-port="5" to-layer="2453" to-port="0" />
+		<edge from-layer="2452" from-port="5" to-layer="2457" to-port="0" />
+		<edge from-layer="2452" from-port="5" to-layer="2463" to-port="0" />
+		<edge from-layer="2453" from-port="2" to-layer="2466" to-port="0" />
+		<edge from-layer="2454" from-port="0" to-layer="2457" to-port="1" />
+		<edge from-layer="2455" from-port="0" to-layer="2457" to-port="2" />
+		<edge from-layer="2456" from-port="0" to-layer="2457" to-port="3" />
+		<edge from-layer="2457" from-port="5" to-layer="2459" to-port="0" />
+		<edge from-layer="2458" from-port="0" to-layer="2459" to-port="1" />
+		<edge from-layer="2459" from-port="2" to-layer="2464" to-port="0" />
+		<edge from-layer="2460" from-port="0" to-layer="2463" to-port="1" />
+		<edge from-layer="2461" from-port="0" to-layer="2463" to-port="2" />
+		<edge from-layer="2462" from-port="0" to-layer="2463" to-port="3" />
+		<edge from-layer="2463" from-port="5" to-layer="2464" to-port="1" />
+		<edge from-layer="2464" from-port="2" to-layer="2465" to-port="0" />
+		<edge from-layer="2465" from-port="2" to-layer="2466" to-port="1" />
+		<edge from-layer="2466" from-port="2" to-layer="2471" to-port="0" />
+		<edge from-layer="2467" from-port="0" to-layer="2470" to-port="1" />
+		<edge from-layer="2468" from-port="0" to-layer="2470" to-port="2" />
+		<edge from-layer="2469" from-port="0" to-layer="2470" to-port="3" />
+		<edge from-layer="2470" from-port="5" to-layer="2471" to-port="1" />
+		<edge from-layer="2471" from-port="2" to-layer="2472" to-port="1" />
+		<edge from-layer="2472" from-port="2" to-layer="2497" to-port="1" />
+		<edge from-layer="2472" from-port="2" to-layer="5147" to-port="0" />
+		<edge from-layer="2473" from-port="0" to-layer="2478" to-port="0" />
+		<edge from-layer="2474" from-port="0" to-layer="2477" to-port="1" />
+		<edge from-layer="2475" from-port="0" to-layer="2477" to-port="2" />
+		<edge from-layer="2476" from-port="0" to-layer="2477" to-port="3" />
+		<edge from-layer="2477" from-port="4" to-layer="2478" to-port="1" />
+		<edge from-layer="2478" from-port="2" to-layer="2479" to-port="0" />
+		<edge from-layer="2479" from-port="1" to-layer="2481" to-port="0" />
+		<edge from-layer="2480" from-port="0" to-layer="2481" to-port="2" />
+		<edge from-layer="2481" from-port="3" to-layer="2488" to-port="0" />
+		<edge from-layer="2482" from-port="0" to-layer="2485" to-port="1" />
+		<edge from-layer="2483" from-port="0" to-layer="2485" to-port="2" />
+		<edge from-layer="2484" from-port="0" to-layer="2485" to-port="3" />
+		<edge from-layer="2485" from-port="5" to-layer="2487" to-port="0" />
+		<edge from-layer="2486" from-port="0" to-layer="2487" to-port="1" />
+		<edge from-layer="2487" from-port="2" to-layer="2488" to-port="1" />
+		<edge from-layer="2488" from-port="2" to-layer="2497" to-port="2" />
+		<edge from-layer="2488" from-port="2" to-layer="5146" to-port="0" />
+		<edge from-layer="2489" from-port="0" to-layer="2496" to-port="1" />
+		<edge from-layer="2490" from-port="1" to-layer="2493" to-port="0" />
+		<edge from-layer="2491" from-port="0" to-layer="2493" to-port="1" />
+		<edge from-layer="2492" from-port="0" to-layer="2493" to-port="2" />
+		<edge from-layer="2493" from-port="3" to-layer="2494" to-port="0" />
+		<edge from-layer="2494" from-port="2" to-layer="2496" to-port="2" />
+		<edge from-layer="2495" from-port="0" to-layer="2496" to-port="3" />
+		<edge from-layer="2496" from-port="5" to-layer="2497" to-port="3" />
+		<edge from-layer="2497" from-port="4" to-layer="2499" to-port="0" />
 		<edge from-layer="2498" from-port="0" to-layer="2499" to-port="1" />
-		<edge from-layer="2499" from-port="2" to-layer="2500" to-port="0" />
-		<edge from-layer="2500" from-port="1" to-layer="2501" to-port="1" />
-		<edge from-layer="2501" from-port="2" to-layer="2504" to-port="0" />
+		<edge from-layer="2499" from-port="2" to-layer="2501" to-port="0" />
+		<edge from-layer="2500" from-port="0" to-layer="2501" to-port="1" />
+		<edge from-layer="2501" from-port="2" to-layer="2510" to-port="0" />
 		<edge from-layer="2502" from-port="0" to-layer="2503" to-port="0" />
-		<edge from-layer="2503" from-port="1" to-layer="2504" to-port="1" />
-		<edge from-layer="2504" from-port="2" to-layer="2505" to-port="0" />
-		<edge from-layer="2505" from-port="1" to-layer="2514" to-port="0" />
-		<edge from-layer="2506" from-port="0" to-layer="2507" to-port="0" />
-		<edge from-layer="2507" from-port="1" to-layer="2510" to-port="0" />
-		<edge from-layer="2508" from-port="0" to-layer="2509" to-port="0" />
+		<edge from-layer="2503" from-port="1" to-layer="2506" to-port="0" />
+		<edge from-layer="2504" from-port="0" to-layer="2505" to-port="0" />
+		<edge from-layer="2505" from-port="1" to-layer="2506" to-port="1" />
+		<edge from-layer="2506" from-port="2" to-layer="2508" to-port="0" />
+		<edge from-layer="2507" from-port="0" to-layer="2508" to-port="1" />
+		<edge from-layer="2508" from-port="2" to-layer="2509" to-port="0" />
 		<edge from-layer="2509" from-port="1" to-layer="2510" to-port="1" />
 		<edge from-layer="2510" from-port="2" to-layer="2512" to-port="0" />
 		<edge from-layer="2511" from-port="0" to-layer="2512" to-port="1" />
-		<edge from-layer="2512" from-port="2" to-layer="2513" to-port="0" />
-		<edge from-layer="2513" from-port="1" to-layer="2514" to-port="1" />
-		<edge from-layer="2514" from-port="2" to-layer="2517" to-port="0" />
-		<edge from-layer="2515" from-port="0" to-layer="2516" to-port="0" />
-		<edge from-layer="2516" from-port="1" to-layer="2517" to-port="1" />
-		<edge from-layer="2517" from-port="2" to-layer="2725" to-port="0" />
-		<edge from-layer="2518" from-port="0" to-layer="2519" to-port="0" />
-		<edge from-layer="2519" from-port="1" to-layer="2520" to-port="1" />
-		<edge from-layer="2520" from-port="2" to-layer="2523" to-port="0" />
+		<edge from-layer="2512" from-port="2" to-layer="2513" to-port="1" />
+		<edge from-layer="2513" from-port="2" to-layer="2514" to-port="0" />
+		<edge from-layer="2514" from-port="2" to-layer="2516" to-port="0" />
+		<edge from-layer="2514" from-port="2" to-layer="2672" to-port="1" />
+		<edge from-layer="2515" from-port="0" to-layer="2516" to-port="1" />
+		<edge from-layer="2516" from-port="2" to-layer="2518" to-port="0" />
+		<edge from-layer="2516" from-port="2" to-layer="2547" to-port="0" />
+		<edge from-layer="2517" from-port="0" to-layer="2518" to-port="1" />
+		<edge from-layer="2518" from-port="2" to-layer="2520" to-port="0" />
+		<edge from-layer="2519" from-port="0" to-layer="2520" to-port="1" />
+		<edge from-layer="2520" from-port="2" to-layer="2529" to-port="0" />
 		<edge from-layer="2521" from-port="0" to-layer="2522" to-port="0" />
-		<edge from-layer="2522" from-port="1" to-layer="2523" to-port="1" />
-		<edge from-layer="2523" from-port="2" to-layer="2532" to-port="0" />
-		<edge from-layer="2524" from-port="0" to-layer="2525" to-port="0" />
-		<edge from-layer="2525" from-port="1" to-layer="2528" to-port="0" />
-		<edge from-layer="2526" from-port="0" to-layer="2527" to-port="0" />
-		<edge from-layer="2527" from-port="1" to-layer="2528" to-port="1" />
-		<edge from-layer="2528" from-port="2" to-layer="2530" to-port="0" />
-		<edge from-layer="2529" from-port="0" to-layer="2530" to-port="1" />
-		<edge from-layer="2530" from-port="2" to-layer="2531" to-port="0" />
+		<edge from-layer="2522" from-port="1" to-layer="2525" to-port="0" />
+		<edge from-layer="2523" from-port="0" to-layer="2524" to-port="0" />
+		<edge from-layer="2524" from-port="1" to-layer="2525" to-port="1" />
+		<edge from-layer="2525" from-port="2" to-layer="2527" to-port="0" />
+		<edge from-layer="2526" from-port="0" to-layer="2527" to-port="1" />
+		<edge from-layer="2527" from-port="2" to-layer="2528" to-port="0" />
+		<edge from-layer="2528" from-port="1" to-layer="2529" to-port="1" />
+		<edge from-layer="2529" from-port="2" to-layer="2532" to-port="0" />
+		<edge from-layer="2530" from-port="0" to-layer="2531" to-port="0" />
 		<edge from-layer="2531" from-port="1" to-layer="2532" to-port="1" />
-		<edge from-layer="2532" from-port="2" to-layer="2535" to-port="0" />
-		<edge from-layer="2533" from-port="0" to-layer="2534" to-port="0" />
-		<edge from-layer="2534" from-port="1" to-layer="2535" to-port="1" />
-		<edge from-layer="2535" from-port="2" to-layer="2537" to-port="0" />
-		<edge from-layer="2536" from-port="0" to-layer="2537" to-port="1" />
-		<edge from-layer="2537" from-port="2" to-layer="2541" to-port="0" />
-		<edge from-layer="2537" from-port="2" to-layer="2665" to-port="0" />
-		<edge from-layer="2537" from-port="2" to-layer="2554" to-port="0" />
-		<edge from-layer="2538" from-port="0" to-layer="2541" to-port="1" />
-		<edge from-layer="2539" from-port="0" to-layer="2541" to-port="2" />
-		<edge from-layer="2540" from-port="0" to-layer="2541" to-port="3" />
-		<edge from-layer="2541" from-port="4" to-layer="2543" to-port="0" />
-		<edge from-layer="2542" from-port="0" to-layer="2543" to-port="1" />
-		<edge from-layer="2543" from-port="2" to-layer="2616" to-port="0" />
-		<edge from-layer="2543" from-port="2" to-layer="2547" to-port="0" />
-		<edge from-layer="2544" from-port="0" to-layer="2547" to-port="1" />
-		<edge from-layer="2545" from-port="0" to-layer="2547" to-port="2" />
-		<edge from-layer="2546" from-port="0" to-layer="2547" to-port="3" />
-		<edge from-layer="2547" from-port="4" to-layer="2581" to-port="0" />
-		<edge from-layer="2547" from-port="4" to-layer="2584" to-port="0" />
-		<edge from-layer="2547" from-port="4" to-layer="2601" to-port="0" />
-		<edge from-layer="2547" from-port="4" to-layer="2595" to-port="0" />
-		<edge from-layer="2548" from-port="0" to-layer="2549" to-port="0" />
-		<edge from-layer="2549" from-port="1" to-layer="2577" to-port="0" />
-		<edge from-layer="2550" from-port="0" to-layer="2577" to-port="1" />
-		<edge from-layer="2551" from-port="0" to-layer="2554" to-port="1" />
-		<edge from-layer="2552" from-port="0" to-layer="2554" to-port="2" />
-		<edge from-layer="2553" from-port="0" to-layer="2554" to-port="3" />
-		<edge from-layer="2554" from-port="4" to-layer="2556" to-port="0" />
+		<edge from-layer="2532" from-port="2" to-layer="2533" to-port="0" />
+		<edge from-layer="2533" from-port="1" to-layer="2542" to-port="0" />
+		<edge from-layer="2534" from-port="0" to-layer="2535" to-port="0" />
+		<edge from-layer="2535" from-port="1" to-layer="2538" to-port="0" />
+		<edge from-layer="2536" from-port="0" to-layer="2537" to-port="0" />
+		<edge from-layer="2537" from-port="1" to-layer="2538" to-port="1" />
+		<edge from-layer="2538" from-port="2" to-layer="2540" to-port="0" />
+		<edge from-layer="2539" from-port="0" to-layer="2540" to-port="1" />
+		<edge from-layer="2540" from-port="2" to-layer="2541" to-port="0" />
+		<edge from-layer="2541" from-port="1" to-layer="2542" to-port="1" />
+		<edge from-layer="2542" from-port="2" to-layer="2545" to-port="0" />
+		<edge from-layer="2543" from-port="0" to-layer="2544" to-port="0" />
+		<edge from-layer="2544" from-port="1" to-layer="2545" to-port="1" />
+		<edge from-layer="2545" from-port="2" to-layer="2671" to-port="0" />
+		<edge from-layer="2546" from-port="0" to-layer="2547" to-port="1" />
+		<edge from-layer="2547" from-port="2" to-layer="2549" to-port="0" />
+		<edge from-layer="2548" from-port="0" to-layer="2549" to-port="1" />
+		<edge from-layer="2549" from-port="2" to-layer="2558" to-port="0" />
+		<edge from-layer="2550" from-port="0" to-layer="2551" to-port="0" />
+		<edge from-layer="2551" from-port="1" to-layer="2554" to-port="0" />
+		<edge from-layer="2552" from-port="0" to-layer="2553" to-port="0" />
+		<edge from-layer="2553" from-port="1" to-layer="2554" to-port="1" />
+		<edge from-layer="2554" from-port="2" to-layer="2556" to-port="0" />
 		<edge from-layer="2555" from-port="0" to-layer="2556" to-port="1" />
-		<edge from-layer="2556" from-port="2" to-layer="2649" to-port="0" />
 		<edge from-layer="2556" from-port="2" to-layer="2557" to-port="0" />
-		<edge from-layer="2556" from-port="2" to-layer="2621" to-port="0" />
-		<edge from-layer="2557" from-port="1" to-layer="2559" to-port="0" />
-		<edge from-layer="2558" from-port="0" to-layer="2559" to-port="2" />
-		<edge from-layer="2559" from-port="3" to-layer="2573" to-port="0" />
-		<edge from-layer="2560" from-port="0" to-layer="2561" to-port="0" />
-		<edge from-layer="2561" from-port="1" to-layer="2566" to-port="0" />
-		<edge from-layer="2562" from-port="0" to-layer="2565" to-port="1" />
-		<edge from-layer="2563" from-port="0" to-layer="2565" to-port="2" />
-		<edge from-layer="2564" from-port="0" to-layer="2565" to-port="3" />
-		<edge from-layer="2565" from-port="4" to-layer="2566" to-port="1" />
-		<edge from-layer="2566" from-port="2" to-layer="2567" to-port="0" />
-		<edge from-layer="2567" from-port="1" to-layer="2569" to-port="0" />
-		<edge from-layer="2568" from-port="0" to-layer="2569" to-port="2" />
-		<edge from-layer="2569" from-port="3" to-layer="2651" to-port="0" />
-		<edge from-layer="2569" from-port="3" to-layer="2570" to-port="0" />
-		<edge from-layer="2570" from-port="1" to-layer="2572" to-port="0" />
-		<edge from-layer="2571" from-port="0" to-layer="2572" to-port="2" />
-		<edge from-layer="2572" from-port="3" to-layer="2573" to-port="1" />
-		<edge from-layer="2573" from-port="2" to-layer="2575" to-port="0" />
-		<edge from-layer="2574" from-port="0" to-layer="2575" to-port="1" />
-		<edge from-layer="2575" from-port="2" to-layer="2607" to-port="2" />
-		<edge from-layer="2575" from-port="2" to-layer="2577" to-port="2" />
+		<edge from-layer="2557" from-port="1" to-layer="2558" to-port="1" />
+		<edge from-layer="2558" from-port="2" to-layer="2560" to-port="0" />
+		<edge from-layer="2559" from-port="0" to-layer="2560" to-port="1" />
+		<edge from-layer="2560" from-port="2" to-layer="2562" to-port="0" />
+		<edge from-layer="2561" from-port="0" to-layer="2562" to-port="1" />
+		<edge from-layer="2562" from-port="2" to-layer="2604" to-port="0" />
+		<edge from-layer="2562" from-port="2" to-layer="2643" to-port="0" />
+		<edge from-layer="2562" from-port="2" to-layer="2566" to-port="0" />
+		<edge from-layer="2563" from-port="0" to-layer="2566" to-port="1" />
+		<edge from-layer="2564" from-port="0" to-layer="2566" to-port="2" />
+		<edge from-layer="2565" from-port="0" to-layer="2566" to-port="3" />
+		<edge from-layer="2566" from-port="5" to-layer="2568" to-port="0" />
+		<edge from-layer="2567" from-port="0" to-layer="2568" to-port="1" />
+		<edge from-layer="2568" from-port="2" to-layer="2572" to-port="0" />
+		<edge from-layer="2568" from-port="2" to-layer="2590" to-port="0" />
+		<edge from-layer="2569" from-port="0" to-layer="2572" to-port="1" />
+		<edge from-layer="2570" from-port="0" to-layer="2572" to-port="2" />
+		<edge from-layer="2571" from-port="0" to-layer="2572" to-port="3" />
+		<edge from-layer="2572" from-port="5" to-layer="2577" to-port="0" />
+		<edge from-layer="2572" from-port="5" to-layer="2583" to-port="0" />
+		<edge from-layer="2572" from-port="5" to-layer="2573" to-port="0" />
+		<edge from-layer="2573" from-port="2" to-layer="2586" to-port="0" />
+		<edge from-layer="2574" from-port="0" to-layer="2577" to-port="1" />
+		<edge from-layer="2575" from-port="0" to-layer="2577" to-port="2" />
 		<edge from-layer="2576" from-port="0" to-layer="2577" to-port="3" />
-		<edge from-layer="2577" from-port="4" to-layer="2579" to-port="0" />
-		<edge from-layer="2578" from-port="0" to-layer="2579" to-port="2" />
-		<edge from-layer="2579" from-port="3" to-layer="2580" to-port="0" />
-		<edge from-layer="2580" from-port="2" to-layer="2581" to-port="1" />
-		<edge from-layer="2580" from-port="2" to-layer="2622" to-port="1" />
-		<edge from-layer="2581" from-port="2" to-layer="2612" to-port="0" />
-		<edge from-layer="2582" from-port="0" to-layer="2592" to-port="0" />
-		<edge from-layer="2583" from-port="0" to-layer="2592" to-port="1" />
-		<edge from-layer="2584" from-port="1" to-layer="2586" to-port="0" />
-		<edge from-layer="2585" from-port="0" to-layer="2586" to-port="2" />
-		<edge from-layer="2586" from-port="3" to-layer="2587" to-port="0" />
-		<edge from-layer="2587" from-port="2" to-layer="2588" to-port="0" />
-		<edge from-layer="2588" from-port="1" to-layer="2590" to-port="0" />
-		<edge from-layer="2589" from-port="0" to-layer="2590" to-port="1" />
-		<edge from-layer="2590" from-port="2" to-layer="2592" to-port="2" />
-		<edge from-layer="2591" from-port="0" to-layer="2592" to-port="3" />
-		<edge from-layer="2592" from-port="4" to-layer="2595" to-port="1" />
-		<edge from-layer="2592" from-port="4" to-layer="2601" to-port="2" />
-		<edge from-layer="2593" from-port="0" to-layer="2595" to-port="2" />
-		<edge from-layer="2594" from-port="0" to-layer="2595" to-port="3" />
-		<edge from-layer="2595" from-port="4" to-layer="2598" to-port="0" />
-		<edge from-layer="2596" from-port="0" to-layer="2597" to-port="0" />
-		<edge from-layer="2597" from-port="1" to-layer="2598" to-port="1" />
-		<edge from-layer="2598" from-port="2" to-layer="2602" to-port="0" />
-		<edge from-layer="2599" from-port="0" to-layer="2601" to-port="1" />
-		<edge from-layer="2600" from-port="0" to-layer="2601" to-port="3" />
-		<edge from-layer="2601" from-port="4" to-layer="2602" to-port="1" />
-		<edge from-layer="2602" from-port="2" to-layer="2611" to-port="0" />
-		<edge from-layer="2603" from-port="0" to-layer="2604" to-port="0" />
-		<edge from-layer="2604" from-port="1" to-layer="2607" to-port="0" />
-		<edge from-layer="2605" from-port="0" to-layer="2607" to-port="1" />
-		<edge from-layer="2606" from-port="0" to-layer="2607" to-port="3" />
-		<edge from-layer="2607" from-port="4" to-layer="2609" to-port="0" />
-		<edge from-layer="2608" from-port="0" to-layer="2609" to-port="2" />
-		<edge from-layer="2609" from-port="3" to-layer="2610" to-port="0" />
-		<edge from-layer="2610" from-port="2" to-layer="2644" to-port="1" />
-		<edge from-layer="2610" from-port="2" to-layer="2611" to-port="1" />
-		<edge from-layer="2611" from-port="2" to-layer="2612" to-port="1" />
-		<edge from-layer="2612" from-port="2" to-layer="2617" to-port="0" />
-		<edge from-layer="2613" from-port="0" to-layer="2616" to-port="1" />
-		<edge from-layer="2614" from-port="0" to-layer="2616" to-port="2" />
-		<edge from-layer="2615" from-port="0" to-layer="2616" to-port="3" />
-		<edge from-layer="2616" from-port="4" to-layer="2617" to-port="1" />
-		<edge from-layer="2617" from-port="2" to-layer="2708" to-port="0" />
-		<edge from-layer="2617" from-port="2" to-layer="2675" to-port="0" />
+		<edge from-layer="2577" from-port="5" to-layer="2579" to-port="0" />
+		<edge from-layer="2578" from-port="0" to-layer="2579" to-port="1" />
+		<edge from-layer="2579" from-port="2" to-layer="2584" to-port="0" />
+		<edge from-layer="2580" from-port="0" to-layer="2583" to-port="1" />
+		<edge from-layer="2581" from-port="0" to-layer="2583" to-port="2" />
+		<edge from-layer="2582" from-port="0" to-layer="2583" to-port="3" />
+		<edge from-layer="2583" from-port="5" to-layer="2584" to-port="1" />
+		<edge from-layer="2584" from-port="2" to-layer="2585" to-port="0" />
+		<edge from-layer="2585" from-port="2" to-layer="2586" to-port="1" />
+		<edge from-layer="2586" from-port="2" to-layer="2591" to-port="0" />
+		<edge from-layer="2587" from-port="0" to-layer="2590" to-port="1" />
+		<edge from-layer="2588" from-port="0" to-layer="2590" to-port="2" />
+		<edge from-layer="2589" from-port="0" to-layer="2590" to-port="3" />
+		<edge from-layer="2590" from-port="5" to-layer="2591" to-port="1" />
+		<edge from-layer="2591" from-port="2" to-layer="2655" to-port="0" />
+		<edge from-layer="2592" from-port="0" to-layer="2597" to-port="0" />
+		<edge from-layer="2593" from-port="0" to-layer="2596" to-port="1" />
+		<edge from-layer="2594" from-port="0" to-layer="2596" to-port="2" />
+		<edge from-layer="2595" from-port="0" to-layer="2596" to-port="3" />
+		<edge from-layer="2596" from-port="4" to-layer="2597" to-port="1" />
+		<edge from-layer="2597" from-port="2" to-layer="2598" to-port="0" />
+		<edge from-layer="2598" from-port="1" to-layer="2600" to-port="0" />
+		<edge from-layer="2599" from-port="0" to-layer="2600" to-port="2" />
+		<edge from-layer="2600" from-port="3" to-layer="2630" to-port="0" />
+		<edge from-layer="2600" from-port="3" to-layer="2648" to-port="0" />
+		<edge from-layer="2601" from-port="0" to-layer="2604" to-port="1" />
+		<edge from-layer="2602" from-port="0" to-layer="2604" to-port="2" />
+		<edge from-layer="2603" from-port="0" to-layer="2604" to-port="3" />
+		<edge from-layer="2604" from-port="5" to-layer="2606" to-port="0" />
+		<edge from-layer="2605" from-port="0" to-layer="2606" to-port="1" />
+		<edge from-layer="2606" from-port="2" to-layer="2628" to-port="0" />
+		<edge from-layer="2606" from-port="2" to-layer="2610" to-port="0" />
+		<edge from-layer="2607" from-port="0" to-layer="2610" to-port="1" />
+		<edge from-layer="2608" from-port="0" to-layer="2610" to-port="2" />
+		<edge from-layer="2609" from-port="0" to-layer="2610" to-port="3" />
+		<edge from-layer="2610" from-port="5" to-layer="2611" to-port="0" />
+		<edge from-layer="2610" from-port="5" to-layer="2621" to-port="0" />
+		<edge from-layer="2610" from-port="5" to-layer="2615" to-port="0" />
+		<edge from-layer="2611" from-port="2" to-layer="2624" to-port="0" />
+		<edge from-layer="2612" from-port="0" to-layer="2615" to-port="1" />
+		<edge from-layer="2613" from-port="0" to-layer="2615" to-port="2" />
+		<edge from-layer="2614" from-port="0" to-layer="2615" to-port="3" />
+		<edge from-layer="2615" from-port="5" to-layer="2617" to-port="0" />
+		<edge from-layer="2616" from-port="0" to-layer="2617" to-port="1" />
+		<edge from-layer="2617" from-port="2" to-layer="2622" to-port="0" />
 		<edge from-layer="2618" from-port="0" to-layer="2621" to-port="1" />
 		<edge from-layer="2619" from-port="0" to-layer="2621" to-port="2" />
 		<edge from-layer="2620" from-port="0" to-layer="2621" to-port="3" />
-		<edge from-layer="2621" from-port="4" to-layer="2636" to-port="0" />
-		<edge from-layer="2621" from-port="4" to-layer="2642" to-port="0" />
-		<edge from-layer="2621" from-port="4" to-layer="2625" to-port="0" />
-		<edge from-layer="2621" from-port="4" to-layer="2622" to-port="0" />
-		<edge from-layer="2622" from-port="2" to-layer="2645" to-port="0" />
-		<edge from-layer="2623" from-port="0" to-layer="2633" to-port="0" />
-		<edge from-layer="2624" from-port="0" to-layer="2633" to-port="1" />
-		<edge from-layer="2625" from-port="1" to-layer="2627" to-port="0" />
-		<edge from-layer="2626" from-port="0" to-layer="2627" to-port="2" />
-		<edge from-layer="2627" from-port="3" to-layer="2628" to-port="0" />
-		<edge from-layer="2628" from-port="2" to-layer="2629" to-port="0" />
-		<edge from-layer="2629" from-port="1" to-layer="2631" to-port="0" />
-		<edge from-layer="2630" from-port="0" to-layer="2631" to-port="1" />
-		<edge from-layer="2631" from-port="2" to-layer="2633" to-port="2" />
-		<edge from-layer="2632" from-port="0" to-layer="2633" to-port="3" />
-		<edge from-layer="2633" from-port="4" to-layer="2636" to-port="1" />
-		<edge from-layer="2633" from-port="4" to-layer="2642" to-port="2" />
-		<edge from-layer="2634" from-port="0" to-layer="2636" to-port="2" />
-		<edge from-layer="2635" from-port="0" to-layer="2636" to-port="3" />
-		<edge from-layer="2636" from-port="4" to-layer="2639" to-port="0" />
-		<edge from-layer="2637" from-port="0" to-layer="2638" to-port="0" />
-		<edge from-layer="2638" from-port="1" to-layer="2639" to-port="1" />
-		<edge from-layer="2639" from-port="2" to-layer="2643" to-port="0" />
-		<edge from-layer="2640" from-port="0" to-layer="2642" to-port="1" />
-		<edge from-layer="2641" from-port="0" to-layer="2642" to-port="3" />
-		<edge from-layer="2642" from-port="4" to-layer="2643" to-port="1" />
-		<edge from-layer="2643" from-port="2" to-layer="2644" to-port="0" />
-		<edge from-layer="2644" from-port="2" to-layer="2645" to-port="1" />
-		<edge from-layer="2645" from-port="2" to-layer="2650" to-port="0" />
-		<edge from-layer="2646" from-port="0" to-layer="2649" to-port="1" />
-		<edge from-layer="2647" from-port="0" to-layer="2649" to-port="2" />
-		<edge from-layer="2648" from-port="0" to-layer="2649" to-port="3" />
-		<edge from-layer="2649" from-port="4" to-layer="2650" to-port="1" />
-		<edge from-layer="2650" from-port="2" to-layer="2651" to-port="1" />
-		<edge from-layer="2651" from-port="2" to-layer="2708" to-port="1" />
-		<edge from-layer="2651" from-port="2" to-layer="7718" to-port="0" />
-		<edge from-layer="2651" from-port="2" to-layer="2672" to-port="0" />
-		<edge from-layer="2652" from-port="0" to-layer="2653" to-port="0" />
-		<edge from-layer="2653" from-port="1" to-layer="2658" to-port="0" />
-		<edge from-layer="2654" from-port="0" to-layer="2657" to-port="1" />
-		<edge from-layer="2655" from-port="0" to-layer="2657" to-port="2" />
-		<edge from-layer="2656" from-port="0" to-layer="2657" to-port="3" />
-		<edge from-layer="2657" from-port="4" to-layer="2658" to-port="1" />
-		<edge from-layer="2658" from-port="2" to-layer="2659" to-port="0" />
-		<edge from-layer="2659" from-port="1" to-layer="2661" to-port="0" />
-		<edge from-layer="2660" from-port="0" to-layer="2661" to-port="2" />
-		<edge from-layer="2661" from-port="3" to-layer="2668" to-port="0" />
-		<edge from-layer="2662" from-port="0" to-layer="2665" to-port="1" />
-		<edge from-layer="2663" from-port="0" to-layer="2665" to-port="2" />
-		<edge from-layer="2664" from-port="0" to-layer="2665" to-port="3" />
-		<edge from-layer="2665" from-port="4" to-layer="2667" to-port="0" />
-		<edge from-layer="2666" from-port="0" to-layer="2667" to-port="1" />
-		<edge from-layer="2667" from-port="2" to-layer="2668" to-port="1" />
-		<edge from-layer="2668" from-port="2" to-layer="2708" to-port="2" />
-		<edge from-layer="2668" from-port="2" to-layer="7717" to-port="0" />
-		<edge from-layer="2669" from-port="0" to-layer="2687" to-port="0" />
-		<edge from-layer="2670" from-port="0" to-layer="2682" to-port="0" />
-		<edge from-layer="2671" from-port="0" to-layer="2682" to-port="1" />
-		<edge from-layer="2671" from-port="0" to-layer="2685" to-port="1" />
-		<edge from-layer="2672" from-port="1" to-layer="2674" to-port="0" />
-		<edge from-layer="2673" from-port="0" to-layer="2674" to-port="2" />
-		<edge from-layer="2674" from-port="3" to-layer="2684" to-port="0" />
-		<edge from-layer="2674" from-port="3" to-layer="2678" to-port="0" />
-		<edge from-layer="2675" from-port="1" to-layer="2703" to-port="0" />
-		<edge from-layer="2675" from-port="1" to-layer="2677" to-port="0" />
-		<edge from-layer="2676" from-port="0" to-layer="2677" to-port="2" />
-		<edge from-layer="2677" from-port="3" to-layer="2678" to-port="1" />
-		<edge from-layer="2678" from-port="2" to-layer="2680" to-port="0" />
-		<edge from-layer="2679" from-port="0" to-layer="2684" to-port="1" />
-		<edge from-layer="2679" from-port="0" to-layer="2680" to-port="1" />
-		<edge from-layer="2680" from-port="2" to-layer="2682" to-port="2" />
-		<edge from-layer="2681" from-port="0" to-layer="2682" to-port="3" />
-		<edge from-layer="2681" from-port="0" to-layer="2685" to-port="3" />
-		<edge from-layer="2682" from-port="4" to-layer="2687" to-port="1" />
-		<edge from-layer="2683" from-port="0" to-layer="2685" to-port="0" />
-		<edge from-layer="2684" from-port="2" to-layer="2685" to-port="2" />
-		<edge from-layer="2684" from-port="2" to-layer="2692" to-port="2" />
-		<edge from-layer="2685" from-port="4" to-layer="2687" to-port="2" />
-		<edge from-layer="2686" from-port="0" to-layer="2687" to-port="3" />
-		<edge from-layer="2687" from-port="4" to-layer="2694" to-port="0" />
-		<edge from-layer="2688" from-port="0" to-layer="2694" to-port="1" />
-		<edge from-layer="2689" from-port="0" to-layer="2692" to-port="0" />
-		<edge from-layer="2690" from-port="0" to-layer="2692" to-port="1" />
-		<edge from-layer="2691" from-port="0" to-layer="2692" to-port="3" />
-		<edge from-layer="2692" from-port="4" to-layer="2694" to-port="2" />
-		<edge from-layer="2693" from-port="0" to-layer="2694" to-port="3" />
-		<edge from-layer="2694" from-port="4" to-layer="2698" to-port="0" />
-		<edge from-layer="2695" from-port="0" to-layer="2696" to-port="0" />
-		<edge from-layer="2696" from-port="1" to-layer="2698" to-port="1" />
-		<edge from-layer="2697" from-port="0" to-layer="2698" to-port="2" />
-		<edge from-layer="2698" from-port="3" to-layer="2699" to-port="0" />
-		<edge from-layer="2699" from-port="2" to-layer="2700" to-port="0" />
-		<edge from-layer="2699" from-port="2" to-layer="2707" to-port="0" />
-		<edge from-layer="2700" from-port="1" to-layer="2706" to-port="0" />
-		<edge from-layer="2701" from-port="0" to-layer="2703" to-port="1" />
-		<edge from-layer="2702" from-port="0" to-layer="2703" to-port="2" />
-		<edge from-layer="2703" from-port="3" to-layer="2704" to-port="0" />
-		<edge from-layer="2704" from-port="4" to-layer="2705" to-port="0" />
-		<edge from-layer="2705" from-port="1" to-layer="2706" to-port="1" />
-		<edge from-layer="2706" from-port="2" to-layer="2707" to-port="1" />
-		<edge from-layer="2707" from-port="2" to-layer="2708" to-port="3" />
-		<edge from-layer="2708" from-port="4" to-layer="2710" to-port="0" />
-		<edge from-layer="2709" from-port="0" to-layer="2710" to-port="1" />
-		<edge from-layer="2710" from-port="2" to-layer="2712" to-port="0" />
-		<edge from-layer="2711" from-port="0" to-layer="2712" to-port="1" />
-		<edge from-layer="2712" from-port="2" to-layer="2721" to-port="0" />
-		<edge from-layer="2713" from-port="0" to-layer="2714" to-port="0" />
-		<edge from-layer="2714" from-port="1" to-layer="2717" to-port="0" />
-		<edge from-layer="2715" from-port="0" to-layer="2716" to-port="0" />
-		<edge from-layer="2716" from-port="1" to-layer="2717" to-port="1" />
-		<edge from-layer="2717" from-port="2" to-layer="2719" to-port="0" />
-		<edge from-layer="2718" from-port="0" to-layer="2719" to-port="1" />
-		<edge from-layer="2719" from-port="2" to-layer="2720" to-port="0" />
-		<edge from-layer="2720" from-port="1" to-layer="2721" to-port="1" />
-		<edge from-layer="2721" from-port="2" to-layer="2724" to-port="0" />
-		<edge from-layer="2722" from-port="0" to-layer="2723" to-port="0" />
-		<edge from-layer="2723" from-port="1" to-layer="2724" to-port="1" />
-		<edge from-layer="2724" from-port="2" to-layer="2725" to-port="1" />
-		<edge from-layer="2725" from-port="2" to-layer="2726" to-port="0" />
-		<edge from-layer="2726" from-port="2" to-layer="2728" to-port="0" />
-		<edge from-layer="2726" from-port="2" to-layer="2968" to-port="1" />
-		<edge from-layer="2727" from-port="0" to-layer="2728" to-port="1" />
-		<edge from-layer="2728" from-port="2" to-layer="2731" to-port="0" />
-		<edge from-layer="2728" from-port="2" to-layer="2762" to-port="0" />
-		<edge from-layer="2729" from-port="0" to-layer="2730" to-port="0" />
-		<edge from-layer="2730" from-port="1" to-layer="2731" to-port="1" />
-		<edge from-layer="2731" from-port="2" to-layer="2734" to-port="0" />
-		<edge from-layer="2732" from-port="0" to-layer="2733" to-port="0" />
-		<edge from-layer="2733" from-port="1" to-layer="2734" to-port="1" />
-		<edge from-layer="2734" from-port="2" to-layer="2743" to-port="0" />
-		<edge from-layer="2735" from-port="0" to-layer="2736" to-port="0" />
-		<edge from-layer="2736" from-port="1" to-layer="2739" to-port="0" />
-		<edge from-layer="2737" from-port="0" to-layer="2738" to-port="0" />
-		<edge from-layer="2738" from-port="1" to-layer="2739" to-port="1" />
-		<edge from-layer="2739" from-port="2" to-layer="2741" to-port="0" />
-		<edge from-layer="2740" from-port="0" to-layer="2741" to-port="1" />
-		<edge from-layer="2741" from-port="2" to-layer="2742" to-port="0" />
-		<edge from-layer="2742" from-port="1" to-layer="2743" to-port="1" />
-		<edge from-layer="2743" from-port="2" to-layer="2746" to-port="0" />
-		<edge from-layer="2744" from-port="0" to-layer="2745" to-port="0" />
-		<edge from-layer="2745" from-port="1" to-layer="2746" to-port="1" />
-		<edge from-layer="2746" from-port="2" to-layer="2747" to-port="0" />
-		<edge from-layer="2747" from-port="1" to-layer="2756" to-port="0" />
-		<edge from-layer="2748" from-port="0" to-layer="2749" to-port="0" />
-		<edge from-layer="2749" from-port="1" to-layer="2752" to-port="0" />
-		<edge from-layer="2750" from-port="0" to-layer="2751" to-port="0" />
-		<edge from-layer="2751" from-port="1" to-layer="2752" to-port="1" />
-		<edge from-layer="2752" from-port="2" to-layer="2754" to-port="0" />
-		<edge from-layer="2753" from-port="0" to-layer="2754" to-port="1" />
-		<edge from-layer="2754" from-port="2" to-layer="2755" to-port="0" />
-		<edge from-layer="2755" from-port="1" to-layer="2756" to-port="1" />
-		<edge from-layer="2756" from-port="2" to-layer="2759" to-port="0" />
-		<edge from-layer="2757" from-port="0" to-layer="2758" to-port="0" />
-		<edge from-layer="2758" from-port="1" to-layer="2759" to-port="1" />
-		<edge from-layer="2759" from-port="2" to-layer="2967" to-port="0" />
-		<edge from-layer="2760" from-port="0" to-layer="2761" to-port="0" />
-		<edge from-layer="2761" from-port="1" to-layer="2762" to-port="1" />
-		<edge from-layer="2762" from-port="2" to-layer="2765" to-port="0" />
-		<edge from-layer="2763" from-port="0" to-layer="2764" to-port="0" />
-		<edge from-layer="2764" from-port="1" to-layer="2765" to-port="1" />
-		<edge from-layer="2765" from-port="2" to-layer="2774" to-port="0" />
-		<edge from-layer="2766" from-port="0" to-layer="2767" to-port="0" />
-		<edge from-layer="2767" from-port="1" to-layer="2770" to-port="0" />
-		<edge from-layer="2768" from-port="0" to-layer="2769" to-port="0" />
-		<edge from-layer="2769" from-port="1" to-layer="2770" to-port="1" />
-		<edge from-layer="2770" from-port="2" to-layer="2772" to-port="0" />
-		<edge from-layer="2771" from-port="0" to-layer="2772" to-port="1" />
-		<edge from-layer="2772" from-port="2" to-layer="2773" to-port="0" />
-		<edge from-layer="2773" from-port="1" to-layer="2774" to-port="1" />
-		<edge from-layer="2774" from-port="2" to-layer="2777" to-port="0" />
-		<edge from-layer="2775" from-port="0" to-layer="2776" to-port="0" />
-		<edge from-layer="2776" from-port="1" to-layer="2777" to-port="1" />
-		<edge from-layer="2777" from-port="2" to-layer="2779" to-port="0" />
-		<edge from-layer="2778" from-port="0" to-layer="2779" to-port="1" />
-		<edge from-layer="2779" from-port="2" to-layer="2796" to-port="0" />
-		<edge from-layer="2779" from-port="2" to-layer="2783" to-port="0" />
-		<edge from-layer="2779" from-port="2" to-layer="2907" to-port="0" />
-		<edge from-layer="2780" from-port="0" to-layer="2783" to-port="1" />
-		<edge from-layer="2781" from-port="0" to-layer="2783" to-port="2" />
-		<edge from-layer="2782" from-port="0" to-layer="2783" to-port="3" />
-		<edge from-layer="2783" from-port="4" to-layer="2785" to-port="0" />
-		<edge from-layer="2784" from-port="0" to-layer="2785" to-port="1" />
-		<edge from-layer="2785" from-port="2" to-layer="2789" to-port="0" />
-		<edge from-layer="2785" from-port="2" to-layer="2858" to-port="0" />
-		<edge from-layer="2786" from-port="0" to-layer="2789" to-port="1" />
-		<edge from-layer="2787" from-port="0" to-layer="2789" to-port="2" />
-		<edge from-layer="2788" from-port="0" to-layer="2789" to-port="3" />
-		<edge from-layer="2789" from-port="4" to-layer="2823" to-port="0" />
-		<edge from-layer="2789" from-port="4" to-layer="2837" to-port="0" />
-		<edge from-layer="2789" from-port="4" to-layer="2843" to-port="0" />
-		<edge from-layer="2789" from-port="4" to-layer="2826" to-port="0" />
-		<edge from-layer="2790" from-port="0" to-layer="2791" to-port="0" />
-		<edge from-layer="2791" from-port="1" to-layer="2819" to-port="0" />
-		<edge from-layer="2792" from-port="0" to-layer="2819" to-port="1" />
-		<edge from-layer="2793" from-port="0" to-layer="2796" to-port="1" />
-		<edge from-layer="2794" from-port="0" to-layer="2796" to-port="2" />
-		<edge from-layer="2795" from-port="0" to-layer="2796" to-port="3" />
-		<edge from-layer="2796" from-port="4" to-layer="2798" to-port="0" />
-		<edge from-layer="2797" from-port="0" to-layer="2798" to-port="1" />
-		<edge from-layer="2798" from-port="2" to-layer="2863" to-port="0" />
-		<edge from-layer="2798" from-port="2" to-layer="2799" to-port="0" />
-		<edge from-layer="2798" from-port="2" to-layer="2891" to-port="0" />
-		<edge from-layer="2799" from-port="1" to-layer="2801" to-port="0" />
-		<edge from-layer="2800" from-port="0" to-layer="2801" to-port="2" />
-		<edge from-layer="2801" from-port="3" to-layer="2815" to-port="0" />
-		<edge from-layer="2802" from-port="0" to-layer="2803" to-port="0" />
-		<edge from-layer="2803" from-port="1" to-layer="2808" to-port="0" />
-		<edge from-layer="2804" from-port="0" to-layer="2807" to-port="1" />
-		<edge from-layer="2805" from-port="0" to-layer="2807" to-port="2" />
-		<edge from-layer="2806" from-port="0" to-layer="2807" to-port="3" />
-		<edge from-layer="2807" from-port="4" to-layer="2808" to-port="1" />
-		<edge from-layer="2808" from-port="2" to-layer="2809" to-port="0" />
-		<edge from-layer="2809" from-port="1" to-layer="2811" to-port="0" />
-		<edge from-layer="2810" from-port="0" to-layer="2811" to-port="2" />
-		<edge from-layer="2811" from-port="3" to-layer="2893" to-port="0" />
-		<edge from-layer="2811" from-port="3" to-layer="2812" to-port="0" />
-		<edge from-layer="2812" from-port="1" to-layer="2814" to-port="0" />
-		<edge from-layer="2813" from-port="0" to-layer="2814" to-port="2" />
-		<edge from-layer="2814" from-port="3" to-layer="2815" to-port="1" />
+		<edge from-layer="2621" from-port="5" to-layer="2622" to-port="1" />
+		<edge from-layer="2622" from-port="2" to-layer="2623" to-port="0" />
+		<edge from-layer="2623" from-port="2" to-layer="2624" to-port="1" />
+		<edge from-layer="2624" from-port="2" to-layer="2629" to-port="0" />
+		<edge from-layer="2625" from-port="0" to-layer="2628" to-port="1" />
+		<edge from-layer="2626" from-port="0" to-layer="2628" to-port="2" />
+		<edge from-layer="2627" from-port="0" to-layer="2628" to-port="3" />
+		<edge from-layer="2628" from-port="5" to-layer="2629" to-port="1" />
+		<edge from-layer="2629" from-port="2" to-layer="2630" to-port="1" />
+		<edge from-layer="2630" from-port="2" to-layer="5145" to-port="0" />
+		<edge from-layer="2630" from-port="2" to-layer="2655" to-port="1" />
+		<edge from-layer="2631" from-port="0" to-layer="2636" to-port="0" />
+		<edge from-layer="2632" from-port="0" to-layer="2635" to-port="1" />
+		<edge from-layer="2633" from-port="0" to-layer="2635" to-port="2" />
+		<edge from-layer="2634" from-port="0" to-layer="2635" to-port="3" />
+		<edge from-layer="2635" from-port="4" to-layer="2636" to-port="1" />
+		<edge from-layer="2636" from-port="2" to-layer="2637" to-port="0" />
+		<edge from-layer="2637" from-port="1" to-layer="2639" to-port="0" />
+		<edge from-layer="2638" from-port="0" to-layer="2639" to-port="2" />
+		<edge from-layer="2639" from-port="3" to-layer="2646" to-port="0" />
+		<edge from-layer="2640" from-port="0" to-layer="2643" to-port="1" />
+		<edge from-layer="2641" from-port="0" to-layer="2643" to-port="2" />
+		<edge from-layer="2642" from-port="0" to-layer="2643" to-port="3" />
+		<edge from-layer="2643" from-port="5" to-layer="2645" to-port="0" />
+		<edge from-layer="2644" from-port="0" to-layer="2645" to-port="1" />
+		<edge from-layer="2645" from-port="2" to-layer="2646" to-port="1" />
+		<edge from-layer="2646" from-port="2" to-layer="5144" to-port="0" />
+		<edge from-layer="2646" from-port="2" to-layer="2655" to-port="2" />
+		<edge from-layer="2647" from-port="0" to-layer="2654" to-port="1" />
+		<edge from-layer="2648" from-port="1" to-layer="2651" to-port="0" />
+		<edge from-layer="2649" from-port="0" to-layer="2651" to-port="1" />
+		<edge from-layer="2650" from-port="0" to-layer="2651" to-port="2" />
+		<edge from-layer="2651" from-port="3" to-layer="2652" to-port="0" />
+		<edge from-layer="2652" from-port="2" to-layer="2654" to-port="2" />
+		<edge from-layer="2653" from-port="0" to-layer="2654" to-port="3" />
+		<edge from-layer="2654" from-port="5" to-layer="2655" to-port="3" />
+		<edge from-layer="2655" from-port="4" to-layer="2657" to-port="0" />
+		<edge from-layer="2656" from-port="0" to-layer="2657" to-port="1" />
+		<edge from-layer="2657" from-port="2" to-layer="2659" to-port="0" />
+		<edge from-layer="2658" from-port="0" to-layer="2659" to-port="1" />
+		<edge from-layer="2659" from-port="2" to-layer="2668" to-port="0" />
+		<edge from-layer="2660" from-port="0" to-layer="2661" to-port="0" />
+		<edge from-layer="2661" from-port="1" to-layer="2664" to-port="0" />
+		<edge from-layer="2662" from-port="0" to-layer="2663" to-port="0" />
+		<edge from-layer="2663" from-port="1" to-layer="2664" to-port="1" />
+		<edge from-layer="2664" from-port="2" to-layer="2666" to-port="0" />
+		<edge from-layer="2665" from-port="0" to-layer="2666" to-port="1" />
+		<edge from-layer="2666" from-port="2" to-layer="2667" to-port="0" />
+		<edge from-layer="2667" from-port="1" to-layer="2668" to-port="1" />
+		<edge from-layer="2668" from-port="2" to-layer="2670" to-port="0" />
+		<edge from-layer="2669" from-port="0" to-layer="2670" to-port="1" />
+		<edge from-layer="2670" from-port="2" to-layer="2671" to-port="1" />
+		<edge from-layer="2671" from-port="2" to-layer="2672" to-port="0" />
+		<edge from-layer="2672" from-port="2" to-layer="2830" to-port="1" />
+		<edge from-layer="2672" from-port="2" to-layer="2674" to-port="0" />
+		<edge from-layer="2673" from-port="0" to-layer="2674" to-port="1" />
+		<edge from-layer="2674" from-port="2" to-layer="2676" to-port="0" />
+		<edge from-layer="2674" from-port="2" to-layer="2705" to-port="0" />
+		<edge from-layer="2675" from-port="0" to-layer="2676" to-port="1" />
+		<edge from-layer="2676" from-port="2" to-layer="2678" to-port="0" />
+		<edge from-layer="2677" from-port="0" to-layer="2678" to-port="1" />
+		<edge from-layer="2678" from-port="2" to-layer="2687" to-port="0" />
+		<edge from-layer="2679" from-port="0" to-layer="2680" to-port="0" />
+		<edge from-layer="2680" from-port="1" to-layer="2683" to-port="0" />
+		<edge from-layer="2681" from-port="0" to-layer="2682" to-port="0" />
+		<edge from-layer="2682" from-port="1" to-layer="2683" to-port="1" />
+		<edge from-layer="2683" from-port="2" to-layer="2685" to-port="0" />
+		<edge from-layer="2684" from-port="0" to-layer="2685" to-port="1" />
+		<edge from-layer="2685" from-port="2" to-layer="2686" to-port="0" />
+		<edge from-layer="2686" from-port="1" to-layer="2687" to-port="1" />
+		<edge from-layer="2687" from-port="2" to-layer="2690" to-port="0" />
+		<edge from-layer="2688" from-port="0" to-layer="2689" to-port="0" />
+		<edge from-layer="2689" from-port="1" to-layer="2690" to-port="1" />
+		<edge from-layer="2690" from-port="2" to-layer="2691" to-port="0" />
+		<edge from-layer="2691" from-port="1" to-layer="2700" to-port="0" />
+		<edge from-layer="2692" from-port="0" to-layer="2693" to-port="0" />
+		<edge from-layer="2693" from-port="1" to-layer="2696" to-port="0" />
+		<edge from-layer="2694" from-port="0" to-layer="2695" to-port="0" />
+		<edge from-layer="2695" from-port="1" to-layer="2696" to-port="1" />
+		<edge from-layer="2696" from-port="2" to-layer="2698" to-port="0" />
+		<edge from-layer="2697" from-port="0" to-layer="2698" to-port="1" />
+		<edge from-layer="2698" from-port="2" to-layer="2699" to-port="0" />
+		<edge from-layer="2699" from-port="1" to-layer="2700" to-port="1" />
+		<edge from-layer="2700" from-port="2" to-layer="2703" to-port="0" />
+		<edge from-layer="2701" from-port="0" to-layer="2702" to-port="0" />
+		<edge from-layer="2702" from-port="1" to-layer="2703" to-port="1" />
+		<edge from-layer="2703" from-port="2" to-layer="2829" to-port="0" />
+		<edge from-layer="2704" from-port="0" to-layer="2705" to-port="1" />
+		<edge from-layer="2705" from-port="2" to-layer="2707" to-port="0" />
+		<edge from-layer="2706" from-port="0" to-layer="2707" to-port="1" />
+		<edge from-layer="2707" from-port="2" to-layer="2716" to-port="0" />
+		<edge from-layer="2708" from-port="0" to-layer="2709" to-port="0" />
+		<edge from-layer="2709" from-port="1" to-layer="2712" to-port="0" />
+		<edge from-layer="2710" from-port="0" to-layer="2711" to-port="0" />
+		<edge from-layer="2711" from-port="1" to-layer="2712" to-port="1" />
+		<edge from-layer="2712" from-port="2" to-layer="2714" to-port="0" />
+		<edge from-layer="2713" from-port="0" to-layer="2714" to-port="1" />
+		<edge from-layer="2714" from-port="2" to-layer="2715" to-port="0" />
+		<edge from-layer="2715" from-port="1" to-layer="2716" to-port="1" />
+		<edge from-layer="2716" from-port="2" to-layer="2718" to-port="0" />
+		<edge from-layer="2717" from-port="0" to-layer="2718" to-port="1" />
+		<edge from-layer="2718" from-port="2" to-layer="2720" to-port="0" />
+		<edge from-layer="2719" from-port="0" to-layer="2720" to-port="1" />
+		<edge from-layer="2720" from-port="2" to-layer="2801" to-port="0" />
+		<edge from-layer="2720" from-port="2" to-layer="2724" to-port="0" />
+		<edge from-layer="2720" from-port="2" to-layer="2762" to-port="0" />
+		<edge from-layer="2721" from-port="0" to-layer="2724" to-port="1" />
+		<edge from-layer="2722" from-port="0" to-layer="2724" to-port="2" />
+		<edge from-layer="2723" from-port="0" to-layer="2724" to-port="3" />
+		<edge from-layer="2724" from-port="5" to-layer="2726" to-port="0" />
+		<edge from-layer="2725" from-port="0" to-layer="2726" to-port="1" />
+		<edge from-layer="2726" from-port="2" to-layer="2730" to-port="0" />
+		<edge from-layer="2726" from-port="2" to-layer="2748" to-port="0" />
+		<edge from-layer="2727" from-port="0" to-layer="2730" to-port="1" />
+		<edge from-layer="2728" from-port="0" to-layer="2730" to-port="2" />
+		<edge from-layer="2729" from-port="0" to-layer="2730" to-port="3" />
+		<edge from-layer="2730" from-port="5" to-layer="2731" to-port="0" />
+		<edge from-layer="2730" from-port="5" to-layer="2735" to-port="0" />
+		<edge from-layer="2730" from-port="5" to-layer="2741" to-port="0" />
+		<edge from-layer="2731" from-port="2" to-layer="2744" to-port="0" />
+		<edge from-layer="2732" from-port="0" to-layer="2735" to-port="1" />
+		<edge from-layer="2733" from-port="0" to-layer="2735" to-port="2" />
+		<edge from-layer="2734" from-port="0" to-layer="2735" to-port="3" />
+		<edge from-layer="2735" from-port="5" to-layer="2737" to-port="0" />
+		<edge from-layer="2736" from-port="0" to-layer="2737" to-port="1" />
+		<edge from-layer="2737" from-port="2" to-layer="2742" to-port="0" />
+		<edge from-layer="2738" from-port="0" to-layer="2741" to-port="1" />
+		<edge from-layer="2739" from-port="0" to-layer="2741" to-port="2" />
+		<edge from-layer="2740" from-port="0" to-layer="2741" to-port="3" />
+		<edge from-layer="2741" from-port="5" to-layer="2742" to-port="1" />
+		<edge from-layer="2742" from-port="2" to-layer="2743" to-port="0" />
+		<edge from-layer="2743" from-port="2" to-layer="2744" to-port="1" />
+		<edge from-layer="2744" from-port="2" to-layer="2749" to-port="0" />
+		<edge from-layer="2745" from-port="0" to-layer="2748" to-port="1" />
+		<edge from-layer="2746" from-port="0" to-layer="2748" to-port="2" />
+		<edge from-layer="2747" from-port="0" to-layer="2748" to-port="3" />
+		<edge from-layer="2748" from-port="5" to-layer="2749" to-port="1" />
+		<edge from-layer="2749" from-port="2" to-layer="2813" to-port="0" />
+		<edge from-layer="2750" from-port="0" to-layer="2755" to-port="0" />
+		<edge from-layer="2751" from-port="0" to-layer="2754" to-port="1" />
+		<edge from-layer="2752" from-port="0" to-layer="2754" to-port="2" />
+		<edge from-layer="2753" from-port="0" to-layer="2754" to-port="3" />
+		<edge from-layer="2754" from-port="4" to-layer="2755" to-port="1" />
+		<edge from-layer="2755" from-port="2" to-layer="2756" to-port="0" />
+		<edge from-layer="2756" from-port="1" to-layer="2758" to-port="0" />
+		<edge from-layer="2757" from-port="0" to-layer="2758" to-port="2" />
+		<edge from-layer="2758" from-port="3" to-layer="2788" to-port="0" />
+		<edge from-layer="2758" from-port="3" to-layer="2806" to-port="0" />
+		<edge from-layer="2759" from-port="0" to-layer="2762" to-port="1" />
+		<edge from-layer="2760" from-port="0" to-layer="2762" to-port="2" />
+		<edge from-layer="2761" from-port="0" to-layer="2762" to-port="3" />
+		<edge from-layer="2762" from-port="5" to-layer="2764" to-port="0" />
+		<edge from-layer="2763" from-port="0" to-layer="2764" to-port="1" />
+		<edge from-layer="2764" from-port="2" to-layer="2786" to-port="0" />
+		<edge from-layer="2764" from-port="2" to-layer="2768" to-port="0" />
+		<edge from-layer="2765" from-port="0" to-layer="2768" to-port="1" />
+		<edge from-layer="2766" from-port="0" to-layer="2768" to-port="2" />
+		<edge from-layer="2767" from-port="0" to-layer="2768" to-port="3" />
+		<edge from-layer="2768" from-port="5" to-layer="2769" to-port="0" />
+		<edge from-layer="2768" from-port="5" to-layer="2773" to-port="0" />
+		<edge from-layer="2768" from-port="5" to-layer="2779" to-port="0" />
+		<edge from-layer="2769" from-port="2" to-layer="2782" to-port="0" />
+		<edge from-layer="2770" from-port="0" to-layer="2773" to-port="1" />
+		<edge from-layer="2771" from-port="0" to-layer="2773" to-port="2" />
+		<edge from-layer="2772" from-port="0" to-layer="2773" to-port="3" />
+		<edge from-layer="2773" from-port="5" to-layer="2775" to-port="0" />
+		<edge from-layer="2774" from-port="0" to-layer="2775" to-port="1" />
+		<edge from-layer="2775" from-port="2" to-layer="2780" to-port="0" />
+		<edge from-layer="2776" from-port="0" to-layer="2779" to-port="1" />
+		<edge from-layer="2777" from-port="0" to-layer="2779" to-port="2" />
+		<edge from-layer="2778" from-port="0" to-layer="2779" to-port="3" />
+		<edge from-layer="2779" from-port="5" to-layer="2780" to-port="1" />
+		<edge from-layer="2780" from-port="2" to-layer="2781" to-port="0" />
+		<edge from-layer="2781" from-port="2" to-layer="2782" to-port="1" />
+		<edge from-layer="2782" from-port="2" to-layer="2787" to-port="0" />
+		<edge from-layer="2783" from-port="0" to-layer="2786" to-port="1" />
+		<edge from-layer="2784" from-port="0" to-layer="2786" to-port="2" />
+		<edge from-layer="2785" from-port="0" to-layer="2786" to-port="3" />
+		<edge from-layer="2786" from-port="5" to-layer="2787" to-port="1" />
+		<edge from-layer="2787" from-port="2" to-layer="2788" to-port="1" />
+		<edge from-layer="2788" from-port="2" to-layer="5143" to-port="0" />
+		<edge from-layer="2788" from-port="2" to-layer="2813" to-port="1" />
+		<edge from-layer="2789" from-port="0" to-layer="2794" to-port="0" />
+		<edge from-layer="2790" from-port="0" to-layer="2793" to-port="1" />
+		<edge from-layer="2791" from-port="0" to-layer="2793" to-port="2" />
+		<edge from-layer="2792" from-port="0" to-layer="2793" to-port="3" />
+		<edge from-layer="2793" from-port="4" to-layer="2794" to-port="1" />
+		<edge from-layer="2794" from-port="2" to-layer="2795" to-port="0" />
+		<edge from-layer="2795" from-port="1" to-layer="2797" to-port="0" />
+		<edge from-layer="2796" from-port="0" to-layer="2797" to-port="2" />
+		<edge from-layer="2797" from-port="3" to-layer="2804" to-port="0" />
+		<edge from-layer="2798" from-port="0" to-layer="2801" to-port="1" />
+		<edge from-layer="2799" from-port="0" to-layer="2801" to-port="2" />
+		<edge from-layer="2800" from-port="0" to-layer="2801" to-port="3" />
+		<edge from-layer="2801" from-port="5" to-layer="2803" to-port="0" />
+		<edge from-layer="2802" from-port="0" to-layer="2803" to-port="1" />
+		<edge from-layer="2803" from-port="2" to-layer="2804" to-port="1" />
+		<edge from-layer="2804" from-port="2" to-layer="2813" to-port="2" />
+		<edge from-layer="2804" from-port="2" to-layer="5142" to-port="0" />
+		<edge from-layer="2805" from-port="0" to-layer="2812" to-port="1" />
+		<edge from-layer="2806" from-port="1" to-layer="2809" to-port="0" />
+		<edge from-layer="2807" from-port="0" to-layer="2809" to-port="1" />
+		<edge from-layer="2808" from-port="0" to-layer="2809" to-port="2" />
+		<edge from-layer="2809" from-port="3" to-layer="2810" to-port="0" />
+		<edge from-layer="2810" from-port="2" to-layer="2812" to-port="2" />
+		<edge from-layer="2811" from-port="0" to-layer="2812" to-port="3" />
+		<edge from-layer="2812" from-port="5" to-layer="2813" to-port="3" />
+		<edge from-layer="2813" from-port="4" to-layer="2815" to-port="0" />
+		<edge from-layer="2814" from-port="0" to-layer="2815" to-port="1" />
 		<edge from-layer="2815" from-port="2" to-layer="2817" to-port="0" />
 		<edge from-layer="2816" from-port="0" to-layer="2817" to-port="1" />
-		<edge from-layer="2817" from-port="2" to-layer="2849" to-port="2" />
-		<edge from-layer="2817" from-port="2" to-layer="2819" to-port="2" />
-		<edge from-layer="2818" from-port="0" to-layer="2819" to-port="3" />
-		<edge from-layer="2819" from-port="4" to-layer="2821" to-port="0" />
-		<edge from-layer="2820" from-port="0" to-layer="2821" to-port="2" />
-		<edge from-layer="2821" from-port="3" to-layer="2822" to-port="0" />
-		<edge from-layer="2822" from-port="2" to-layer="2823" to-port="1" />
-		<edge from-layer="2822" from-port="2" to-layer="2864" to-port="1" />
-		<edge from-layer="2823" from-port="2" to-layer="2854" to-port="0" />
-		<edge from-layer="2824" from-port="0" to-layer="2834" to-port="0" />
-		<edge from-layer="2825" from-port="0" to-layer="2834" to-port="1" />
-		<edge from-layer="2826" from-port="1" to-layer="2828" to-port="0" />
-		<edge from-layer="2827" from-port="0" to-layer="2828" to-port="2" />
-		<edge from-layer="2828" from-port="3" to-layer="2829" to-port="0" />
+		<edge from-layer="2817" from-port="2" to-layer="2826" to-port="0" />
+		<edge from-layer="2818" from-port="0" to-layer="2819" to-port="0" />
+		<edge from-layer="2819" from-port="1" to-layer="2822" to-port="0" />
+		<edge from-layer="2820" from-port="0" to-layer="2821" to-port="0" />
+		<edge from-layer="2821" from-port="1" to-layer="2822" to-port="1" />
+		<edge from-layer="2822" from-port="2" to-layer="2824" to-port="0" />
+		<edge from-layer="2823" from-port="0" to-layer="2824" to-port="1" />
+		<edge from-layer="2824" from-port="2" to-layer="2825" to-port="0" />
+		<edge from-layer="2825" from-port="1" to-layer="2826" to-port="1" />
+		<edge from-layer="2826" from-port="2" to-layer="2828" to-port="0" />
+		<edge from-layer="2827" from-port="0" to-layer="2828" to-port="1" />
+		<edge from-layer="2828" from-port="2" to-layer="2829" to-port="1" />
 		<edge from-layer="2829" from-port="2" to-layer="2830" to-port="0" />
-		<edge from-layer="2830" from-port="1" to-layer="2832" to-port="0" />
+		<edge from-layer="2830" from-port="2" to-layer="2832" to-port="0" />
+		<edge from-layer="2830" from-port="2" to-layer="2988" to-port="1" />
 		<edge from-layer="2831" from-port="0" to-layer="2832" to-port="1" />
-		<edge from-layer="2832" from-port="2" to-layer="2834" to-port="2" />
-		<edge from-layer="2833" from-port="0" to-layer="2834" to-port="3" />
-		<edge from-layer="2834" from-port="4" to-layer="2843" to-port="2" />
-		<edge from-layer="2834" from-port="4" to-layer="2837" to-port="1" />
-		<edge from-layer="2835" from-port="0" to-layer="2837" to-port="2" />
-		<edge from-layer="2836" from-port="0" to-layer="2837" to-port="3" />
-		<edge from-layer="2837" from-port="4" to-layer="2840" to-port="0" />
-		<edge from-layer="2838" from-port="0" to-layer="2839" to-port="0" />
-		<edge from-layer="2839" from-port="1" to-layer="2840" to-port="1" />
-		<edge from-layer="2840" from-port="2" to-layer="2844" to-port="0" />
-		<edge from-layer="2841" from-port="0" to-layer="2843" to-port="1" />
-		<edge from-layer="2842" from-port="0" to-layer="2843" to-port="3" />
-		<edge from-layer="2843" from-port="4" to-layer="2844" to-port="1" />
-		<edge from-layer="2844" from-port="2" to-layer="2853" to-port="0" />
-		<edge from-layer="2845" from-port="0" to-layer="2846" to-port="0" />
-		<edge from-layer="2846" from-port="1" to-layer="2849" to-port="0" />
-		<edge from-layer="2847" from-port="0" to-layer="2849" to-port="1" />
-		<edge from-layer="2848" from-port="0" to-layer="2849" to-port="3" />
-		<edge from-layer="2849" from-port="4" to-layer="2851" to-port="0" />
-		<edge from-layer="2850" from-port="0" to-layer="2851" to-port="2" />
-		<edge from-layer="2851" from-port="3" to-layer="2852" to-port="0" />
-		<edge from-layer="2852" from-port="2" to-layer="2886" to-port="1" />
-		<edge from-layer="2852" from-port="2" to-layer="2853" to-port="1" />
-		<edge from-layer="2853" from-port="2" to-layer="2854" to-port="1" />
-		<edge from-layer="2854" from-port="2" to-layer="2859" to-port="0" />
-		<edge from-layer="2855" from-port="0" to-layer="2858" to-port="1" />
-		<edge from-layer="2856" from-port="0" to-layer="2858" to-port="2" />
-		<edge from-layer="2857" from-port="0" to-layer="2858" to-port="3" />
-		<edge from-layer="2858" from-port="4" to-layer="2859" to-port="1" />
-		<edge from-layer="2859" from-port="2" to-layer="2917" to-port="0" />
-		<edge from-layer="2859" from-port="2" to-layer="2950" to-port="0" />
-		<edge from-layer="2860" from-port="0" to-layer="2863" to-port="1" />
-		<edge from-layer="2861" from-port="0" to-layer="2863" to-port="2" />
-		<edge from-layer="2862" from-port="0" to-layer="2863" to-port="3" />
-		<edge from-layer="2863" from-port="4" to-layer="2878" to-port="0" />
-		<edge from-layer="2863" from-port="4" to-layer="2884" to-port="0" />
-		<edge from-layer="2863" from-port="4" to-layer="2867" to-port="0" />
-		<edge from-layer="2863" from-port="4" to-layer="2864" to-port="0" />
-		<edge from-layer="2864" from-port="2" to-layer="2887" to-port="0" />
-		<edge from-layer="2865" from-port="0" to-layer="2875" to-port="0" />
-		<edge from-layer="2866" from-port="0" to-layer="2875" to-port="1" />
-		<edge from-layer="2867" from-port="1" to-layer="2869" to-port="0" />
-		<edge from-layer="2868" from-port="0" to-layer="2869" to-port="2" />
-		<edge from-layer="2869" from-port="3" to-layer="2870" to-port="0" />
-		<edge from-layer="2870" from-port="2" to-layer="2871" to-port="0" />
-		<edge from-layer="2871" from-port="1" to-layer="2873" to-port="0" />
-		<edge from-layer="2872" from-port="0" to-layer="2873" to-port="1" />
-		<edge from-layer="2873" from-port="2" to-layer="2875" to-port="2" />
-		<edge from-layer="2874" from-port="0" to-layer="2875" to-port="3" />
-		<edge from-layer="2875" from-port="4" to-layer="2884" to-port="2" />
-		<edge from-layer="2875" from-port="4" to-layer="2878" to-port="1" />
-		<edge from-layer="2876" from-port="0" to-layer="2878" to-port="2" />
-		<edge from-layer="2877" from-port="0" to-layer="2878" to-port="3" />
-		<edge from-layer="2878" from-port="4" to-layer="2881" to-port="0" />
-		<edge from-layer="2879" from-port="0" to-layer="2880" to-port="0" />
-		<edge from-layer="2880" from-port="1" to-layer="2881" to-port="1" />
-		<edge from-layer="2881" from-port="2" to-layer="2885" to-port="0" />
-		<edge from-layer="2882" from-port="0" to-layer="2884" to-port="1" />
-		<edge from-layer="2883" from-port="0" to-layer="2884" to-port="3" />
-		<edge from-layer="2884" from-port="4" to-layer="2885" to-port="1" />
-		<edge from-layer="2885" from-port="2" to-layer="2886" to-port="0" />
-		<edge from-layer="2886" from-port="2" to-layer="2887" to-port="1" />
-		<edge from-layer="2887" from-port="2" to-layer="2892" to-port="0" />
-		<edge from-layer="2888" from-port="0" to-layer="2891" to-port="1" />
-		<edge from-layer="2889" from-port="0" to-layer="2891" to-port="2" />
-		<edge from-layer="2890" from-port="0" to-layer="2891" to-port="3" />
-		<edge from-layer="2891" from-port="4" to-layer="2892" to-port="1" />
-		<edge from-layer="2892" from-port="2" to-layer="2893" to-port="1" />
-		<edge from-layer="2893" from-port="2" to-layer="2950" to-port="1" />
-		<edge from-layer="2893" from-port="2" to-layer="2914" to-port="0" />
-		<edge from-layer="2893" from-port="2" to-layer="7716" to-port="0" />
-		<edge from-layer="2894" from-port="0" to-layer="2895" to-port="0" />
-		<edge from-layer="2895" from-port="1" to-layer="2900" to-port="0" />
+		<edge from-layer="2832" from-port="2" to-layer="2863" to-port="0" />
+		<edge from-layer="2832" from-port="2" to-layer="2834" to-port="0" />
+		<edge from-layer="2833" from-port="0" to-layer="2834" to-port="1" />
+		<edge from-layer="2834" from-port="2" to-layer="2836" to-port="0" />
+		<edge from-layer="2835" from-port="0" to-layer="2836" to-port="1" />
+		<edge from-layer="2836" from-port="2" to-layer="2845" to-port="0" />
+		<edge from-layer="2837" from-port="0" to-layer="2838" to-port="0" />
+		<edge from-layer="2838" from-port="1" to-layer="2841" to-port="0" />
+		<edge from-layer="2839" from-port="0" to-layer="2840" to-port="0" />
+		<edge from-layer="2840" from-port="1" to-layer="2841" to-port="1" />
+		<edge from-layer="2841" from-port="2" to-layer="2843" to-port="0" />
+		<edge from-layer="2842" from-port="0" to-layer="2843" to-port="1" />
+		<edge from-layer="2843" from-port="2" to-layer="2844" to-port="0" />
+		<edge from-layer="2844" from-port="1" to-layer="2845" to-port="1" />
+		<edge from-layer="2845" from-port="2" to-layer="2848" to-port="0" />
+		<edge from-layer="2846" from-port="0" to-layer="2847" to-port="0" />
+		<edge from-layer="2847" from-port="1" to-layer="2848" to-port="1" />
+		<edge from-layer="2848" from-port="2" to-layer="2849" to-port="0" />
+		<edge from-layer="2849" from-port="1" to-layer="2858" to-port="0" />
+		<edge from-layer="2850" from-port="0" to-layer="2851" to-port="0" />
+		<edge from-layer="2851" from-port="1" to-layer="2854" to-port="0" />
+		<edge from-layer="2852" from-port="0" to-layer="2853" to-port="0" />
+		<edge from-layer="2853" from-port="1" to-layer="2854" to-port="1" />
+		<edge from-layer="2854" from-port="2" to-layer="2856" to-port="0" />
+		<edge from-layer="2855" from-port="0" to-layer="2856" to-port="1" />
+		<edge from-layer="2856" from-port="2" to-layer="2857" to-port="0" />
+		<edge from-layer="2857" from-port="1" to-layer="2858" to-port="1" />
+		<edge from-layer="2858" from-port="2" to-layer="2861" to-port="0" />
+		<edge from-layer="2859" from-port="0" to-layer="2860" to-port="0" />
+		<edge from-layer="2860" from-port="1" to-layer="2861" to-port="1" />
+		<edge from-layer="2861" from-port="2" to-layer="2987" to-port="0" />
+		<edge from-layer="2862" from-port="0" to-layer="2863" to-port="1" />
+		<edge from-layer="2863" from-port="2" to-layer="2865" to-port="0" />
+		<edge from-layer="2864" from-port="0" to-layer="2865" to-port="1" />
+		<edge from-layer="2865" from-port="2" to-layer="2874" to-port="0" />
+		<edge from-layer="2866" from-port="0" to-layer="2867" to-port="0" />
+		<edge from-layer="2867" from-port="1" to-layer="2870" to-port="0" />
+		<edge from-layer="2868" from-port="0" to-layer="2869" to-port="0" />
+		<edge from-layer="2869" from-port="1" to-layer="2870" to-port="1" />
+		<edge from-layer="2870" from-port="2" to-layer="2872" to-port="0" />
+		<edge from-layer="2871" from-port="0" to-layer="2872" to-port="1" />
+		<edge from-layer="2872" from-port="2" to-layer="2873" to-port="0" />
+		<edge from-layer="2873" from-port="1" to-layer="2874" to-port="1" />
+		<edge from-layer="2874" from-port="2" to-layer="2876" to-port="0" />
+		<edge from-layer="2875" from-port="0" to-layer="2876" to-port="1" />
+		<edge from-layer="2876" from-port="2" to-layer="2878" to-port="0" />
+		<edge from-layer="2877" from-port="0" to-layer="2878" to-port="1" />
+		<edge from-layer="2878" from-port="2" to-layer="2959" to-port="0" />
+		<edge from-layer="2878" from-port="2" to-layer="2882" to-port="0" />
+		<edge from-layer="2878" from-port="2" to-layer="2920" to-port="0" />
+		<edge from-layer="2879" from-port="0" to-layer="2882" to-port="1" />
+		<edge from-layer="2880" from-port="0" to-layer="2882" to-port="2" />
+		<edge from-layer="2881" from-port="0" to-layer="2882" to-port="3" />
+		<edge from-layer="2882" from-port="5" to-layer="2884" to-port="0" />
+		<edge from-layer="2883" from-port="0" to-layer="2884" to-port="1" />
+		<edge from-layer="2884" from-port="2" to-layer="2888" to-port="0" />
+		<edge from-layer="2884" from-port="2" to-layer="2906" to-port="0" />
+		<edge from-layer="2885" from-port="0" to-layer="2888" to-port="1" />
+		<edge from-layer="2886" from-port="0" to-layer="2888" to-port="2" />
+		<edge from-layer="2887" from-port="0" to-layer="2888" to-port="3" />
+		<edge from-layer="2888" from-port="5" to-layer="2889" to-port="0" />
+		<edge from-layer="2888" from-port="5" to-layer="2899" to-port="0" />
+		<edge from-layer="2888" from-port="5" to-layer="2893" to-port="0" />
+		<edge from-layer="2889" from-port="2" to-layer="2902" to-port="0" />
+		<edge from-layer="2890" from-port="0" to-layer="2893" to-port="1" />
+		<edge from-layer="2891" from-port="0" to-layer="2893" to-port="2" />
+		<edge from-layer="2892" from-port="0" to-layer="2893" to-port="3" />
+		<edge from-layer="2893" from-port="5" to-layer="2895" to-port="0" />
+		<edge from-layer="2894" from-port="0" to-layer="2895" to-port="1" />
+		<edge from-layer="2895" from-port="2" to-layer="2900" to-port="0" />
 		<edge from-layer="2896" from-port="0" to-layer="2899" to-port="1" />
 		<edge from-layer="2897" from-port="0" to-layer="2899" to-port="2" />
 		<edge from-layer="2898" from-port="0" to-layer="2899" to-port="3" />
-		<edge from-layer="2899" from-port="4" to-layer="2900" to-port="1" />
+		<edge from-layer="2899" from-port="5" to-layer="2900" to-port="1" />
 		<edge from-layer="2900" from-port="2" to-layer="2901" to-port="0" />
-		<edge from-layer="2901" from-port="1" to-layer="2903" to-port="0" />
-		<edge from-layer="2902" from-port="0" to-layer="2903" to-port="2" />
-		<edge from-layer="2903" from-port="3" to-layer="2910" to-port="0" />
-		<edge from-layer="2904" from-port="0" to-layer="2907" to-port="1" />
-		<edge from-layer="2905" from-port="0" to-layer="2907" to-port="2" />
-		<edge from-layer="2906" from-port="0" to-layer="2907" to-port="3" />
-		<edge from-layer="2907" from-port="4" to-layer="2909" to-port="0" />
-		<edge from-layer="2908" from-port="0" to-layer="2909" to-port="1" />
-		<edge from-layer="2909" from-port="2" to-layer="2910" to-port="1" />
-		<edge from-layer="2910" from-port="2" to-layer="2950" to-port="2" />
-		<edge from-layer="2910" from-port="2" to-layer="7715" to-port="0" />
-		<edge from-layer="2911" from-port="0" to-layer="2929" to-port="0" />
-		<edge from-layer="2912" from-port="0" to-layer="2924" to-port="0" />
-		<edge from-layer="2913" from-port="0" to-layer="2924" to-port="1" />
-		<edge from-layer="2913" from-port="0" to-layer="2927" to-port="1" />
+		<edge from-layer="2901" from-port="2" to-layer="2902" to-port="1" />
+		<edge from-layer="2902" from-port="2" to-layer="2907" to-port="0" />
+		<edge from-layer="2903" from-port="0" to-layer="2906" to-port="1" />
+		<edge from-layer="2904" from-port="0" to-layer="2906" to-port="2" />
+		<edge from-layer="2905" from-port="0" to-layer="2906" to-port="3" />
+		<edge from-layer="2906" from-port="5" to-layer="2907" to-port="1" />
+		<edge from-layer="2907" from-port="2" to-layer="2971" to-port="0" />
+		<edge from-layer="2908" from-port="0" to-layer="2913" to-port="0" />
+		<edge from-layer="2909" from-port="0" to-layer="2912" to-port="1" />
+		<edge from-layer="2910" from-port="0" to-layer="2912" to-port="2" />
+		<edge from-layer="2911" from-port="0" to-layer="2912" to-port="3" />
+		<edge from-layer="2912" from-port="4" to-layer="2913" to-port="1" />
+		<edge from-layer="2913" from-port="2" to-layer="2914" to-port="0" />
 		<edge from-layer="2914" from-port="1" to-layer="2916" to-port="0" />
 		<edge from-layer="2915" from-port="0" to-layer="2916" to-port="2" />
-		<edge from-layer="2916" from-port="3" to-layer="2926" to-port="0" />
-		<edge from-layer="2916" from-port="3" to-layer="2920" to-port="0" />
-		<edge from-layer="2917" from-port="1" to-layer="2919" to-port="0" />
-		<edge from-layer="2917" from-port="1" to-layer="2945" to-port="0" />
-		<edge from-layer="2918" from-port="0" to-layer="2919" to-port="2" />
-		<edge from-layer="2919" from-port="3" to-layer="2920" to-port="1" />
-		<edge from-layer="2920" from-port="2" to-layer="2922" to-port="0" />
-		<edge from-layer="2921" from-port="0" to-layer="2926" to-port="1" />
+		<edge from-layer="2916" from-port="3" to-layer="2964" to-port="0" />
+		<edge from-layer="2916" from-port="3" to-layer="2946" to-port="0" />
+		<edge from-layer="2917" from-port="0" to-layer="2920" to-port="1" />
+		<edge from-layer="2918" from-port="0" to-layer="2920" to-port="2" />
+		<edge from-layer="2919" from-port="0" to-layer="2920" to-port="3" />
+		<edge from-layer="2920" from-port="5" to-layer="2922" to-port="0" />
 		<edge from-layer="2921" from-port="0" to-layer="2922" to-port="1" />
-		<edge from-layer="2922" from-port="2" to-layer="2924" to-port="2" />
-		<edge from-layer="2923" from-port="0" to-layer="2924" to-port="3" />
-		<edge from-layer="2923" from-port="0" to-layer="2927" to-port="3" />
-		<edge from-layer="2924" from-port="4" to-layer="2929" to-port="1" />
-		<edge from-layer="2925" from-port="0" to-layer="2927" to-port="0" />
-		<edge from-layer="2926" from-port="2" to-layer="2934" to-port="2" />
-		<edge from-layer="2926" from-port="2" to-layer="2927" to-port="2" />
-		<edge from-layer="2927" from-port="4" to-layer="2929" to-port="2" />
-		<edge from-layer="2928" from-port="0" to-layer="2929" to-port="3" />
-		<edge from-layer="2929" from-port="4" to-layer="2936" to-port="0" />
-		<edge from-layer="2930" from-port="0" to-layer="2936" to-port="1" />
-		<edge from-layer="2931" from-port="0" to-layer="2934" to-port="0" />
-		<edge from-layer="2932" from-port="0" to-layer="2934" to-port="1" />
-		<edge from-layer="2933" from-port="0" to-layer="2934" to-port="3" />
-		<edge from-layer="2934" from-port="4" to-layer="2936" to-port="2" />
-		<edge from-layer="2935" from-port="0" to-layer="2936" to-port="3" />
-		<edge from-layer="2936" from-port="4" to-layer="2940" to-port="0" />
-		<edge from-layer="2937" from-port="0" to-layer="2938" to-port="0" />
-		<edge from-layer="2938" from-port="1" to-layer="2940" to-port="1" />
-		<edge from-layer="2939" from-port="0" to-layer="2940" to-port="2" />
-		<edge from-layer="2940" from-port="3" to-layer="2941" to-port="0" />
-		<edge from-layer="2941" from-port="2" to-layer="2942" to-port="0" />
-		<edge from-layer="2941" from-port="2" to-layer="2949" to-port="0" />
-		<edge from-layer="2942" from-port="1" to-layer="2948" to-port="0" />
-		<edge from-layer="2943" from-port="0" to-layer="2945" to-port="1" />
-		<edge from-layer="2944" from-port="0" to-layer="2945" to-port="2" />
-		<edge from-layer="2945" from-port="3" to-layer="2946" to-port="0" />
-		<edge from-layer="2946" from-port="4" to-layer="2947" to-port="0" />
-		<edge from-layer="2947" from-port="1" to-layer="2948" to-port="1" />
-		<edge from-layer="2948" from-port="2" to-layer="2949" to-port="1" />
-		<edge from-layer="2949" from-port="2" to-layer="2950" to-port="3" />
-		<edge from-layer="2950" from-port="4" to-layer="2952" to-port="0" />
-		<edge from-layer="2951" from-port="0" to-layer="2952" to-port="1" />
-		<edge from-layer="2952" from-port="2" to-layer="2954" to-port="0" />
-		<edge from-layer="2953" from-port="0" to-layer="2954" to-port="1" />
-		<edge from-layer="2954" from-port="2" to-layer="2963" to-port="0" />
-		<edge from-layer="2955" from-port="0" to-layer="2956" to-port="0" />
-		<edge from-layer="2956" from-port="1" to-layer="2959" to-port="0" />
-		<edge from-layer="2957" from-port="0" to-layer="2958" to-port="0" />
-		<edge from-layer="2958" from-port="1" to-layer="2959" to-port="1" />
-		<edge from-layer="2959" from-port="2" to-layer="2961" to-port="0" />
+		<edge from-layer="2922" from-port="2" to-layer="2926" to-port="0" />
+		<edge from-layer="2922" from-port="2" to-layer="2944" to-port="0" />
+		<edge from-layer="2923" from-port="0" to-layer="2926" to-port="1" />
+		<edge from-layer="2924" from-port="0" to-layer="2926" to-port="2" />
+		<edge from-layer="2925" from-port="0" to-layer="2926" to-port="3" />
+		<edge from-layer="2926" from-port="5" to-layer="2937" to-port="0" />
+		<edge from-layer="2926" from-port="5" to-layer="2931" to-port="0" />
+		<edge from-layer="2926" from-port="5" to-layer="2927" to-port="0" />
+		<edge from-layer="2927" from-port="2" to-layer="2940" to-port="0" />
+		<edge from-layer="2928" from-port="0" to-layer="2931" to-port="1" />
+		<edge from-layer="2929" from-port="0" to-layer="2931" to-port="2" />
+		<edge from-layer="2930" from-port="0" to-layer="2931" to-port="3" />
+		<edge from-layer="2931" from-port="5" to-layer="2933" to-port="0" />
+		<edge from-layer="2932" from-port="0" to-layer="2933" to-port="1" />
+		<edge from-layer="2933" from-port="2" to-layer="2938" to-port="0" />
+		<edge from-layer="2934" from-port="0" to-layer="2937" to-port="1" />
+		<edge from-layer="2935" from-port="0" to-layer="2937" to-port="2" />
+		<edge from-layer="2936" from-port="0" to-layer="2937" to-port="3" />
+		<edge from-layer="2937" from-port="5" to-layer="2938" to-port="1" />
+		<edge from-layer="2938" from-port="2" to-layer="2939" to-port="0" />
+		<edge from-layer="2939" from-port="2" to-layer="2940" to-port="1" />
+		<edge from-layer="2940" from-port="2" to-layer="2945" to-port="0" />
+		<edge from-layer="2941" from-port="0" to-layer="2944" to-port="1" />
+		<edge from-layer="2942" from-port="0" to-layer="2944" to-port="2" />
+		<edge from-layer="2943" from-port="0" to-layer="2944" to-port="3" />
+		<edge from-layer="2944" from-port="5" to-layer="2945" to-port="1" />
+		<edge from-layer="2945" from-port="2" to-layer="2946" to-port="1" />
+		<edge from-layer="2946" from-port="2" to-layer="5141" to-port="0" />
+		<edge from-layer="2946" from-port="2" to-layer="2971" to-port="1" />
+		<edge from-layer="2947" from-port="0" to-layer="2952" to-port="0" />
+		<edge from-layer="2948" from-port="0" to-layer="2951" to-port="1" />
+		<edge from-layer="2949" from-port="0" to-layer="2951" to-port="2" />
+		<edge from-layer="2950" from-port="0" to-layer="2951" to-port="3" />
+		<edge from-layer="2951" from-port="4" to-layer="2952" to-port="1" />
+		<edge from-layer="2952" from-port="2" to-layer="2953" to-port="0" />
+		<edge from-layer="2953" from-port="1" to-layer="2955" to-port="0" />
+		<edge from-layer="2954" from-port="0" to-layer="2955" to-port="2" />
+		<edge from-layer="2955" from-port="3" to-layer="2962" to-port="0" />
+		<edge from-layer="2956" from-port="0" to-layer="2959" to-port="1" />
+		<edge from-layer="2957" from-port="0" to-layer="2959" to-port="2" />
+		<edge from-layer="2958" from-port="0" to-layer="2959" to-port="3" />
+		<edge from-layer="2959" from-port="5" to-layer="2961" to-port="0" />
 		<edge from-layer="2960" from-port="0" to-layer="2961" to-port="1" />
-		<edge from-layer="2961" from-port="2" to-layer="2962" to-port="0" />
-		<edge from-layer="2962" from-port="1" to-layer="2963" to-port="1" />
-		<edge from-layer="2963" from-port="2" to-layer="2966" to-port="0" />
-		<edge from-layer="2964" from-port="0" to-layer="2965" to-port="0" />
-		<edge from-layer="2965" from-port="1" to-layer="2966" to-port="1" />
-		<edge from-layer="2966" from-port="2" to-layer="2967" to-port="1" />
-		<edge from-layer="2967" from-port="2" to-layer="2968" to-port="0" />
-		<edge from-layer="2968" from-port="2" to-layer="2970" to-port="0" />
-		<edge from-layer="2968" from-port="2" to-layer="3210" to-port="1" />
-		<edge from-layer="2969" from-port="0" to-layer="2970" to-port="1" />
-		<edge from-layer="2970" from-port="2" to-layer="2973" to-port="0" />
-		<edge from-layer="2970" from-port="2" to-layer="3004" to-port="0" />
-		<edge from-layer="2971" from-port="0" to-layer="2972" to-port="0" />
-		<edge from-layer="2972" from-port="1" to-layer="2973" to-port="1" />
-		<edge from-layer="2973" from-port="2" to-layer="2976" to-port="0" />
-		<edge from-layer="2974" from-port="0" to-layer="2975" to-port="0" />
-		<edge from-layer="2975" from-port="1" to-layer="2976" to-port="1" />
-		<edge from-layer="2976" from-port="2" to-layer="2985" to-port="0" />
-		<edge from-layer="2977" from-port="0" to-layer="2978" to-port="0" />
-		<edge from-layer="2978" from-port="1" to-layer="2981" to-port="0" />
-		<edge from-layer="2979" from-port="0" to-layer="2980" to-port="0" />
-		<edge from-layer="2980" from-port="1" to-layer="2981" to-port="1" />
-		<edge from-layer="2981" from-port="2" to-layer="2983" to-port="0" />
-		<edge from-layer="2982" from-port="0" to-layer="2983" to-port="1" />
-		<edge from-layer="2983" from-port="2" to-layer="2984" to-port="0" />
-		<edge from-layer="2984" from-port="1" to-layer="2985" to-port="1" />
-		<edge from-layer="2985" from-port="2" to-layer="2988" to-port="0" />
-		<edge from-layer="2986" from-port="0" to-layer="2987" to-port="0" />
-		<edge from-layer="2987" from-port="1" to-layer="2988" to-port="1" />
-		<edge from-layer="2988" from-port="2" to-layer="2989" to-port="0" />
-		<edge from-layer="2989" from-port="1" to-layer="2998" to-port="0" />
-		<edge from-layer="2990" from-port="0" to-layer="2991" to-port="0" />
-		<edge from-layer="2991" from-port="1" to-layer="2994" to-port="0" />
-		<edge from-layer="2992" from-port="0" to-layer="2993" to-port="0" />
-		<edge from-layer="2993" from-port="1" to-layer="2994" to-port="1" />
-		<edge from-layer="2994" from-port="2" to-layer="2996" to-port="0" />
-		<edge from-layer="2995" from-port="0" to-layer="2996" to-port="1" />
-		<edge from-layer="2996" from-port="2" to-layer="2997" to-port="0" />
-		<edge from-layer="2997" from-port="1" to-layer="2998" to-port="1" />
-		<edge from-layer="2998" from-port="2" to-layer="3001" to-port="0" />
-		<edge from-layer="2999" from-port="0" to-layer="3000" to-port="0" />
-		<edge from-layer="3000" from-port="1" to-layer="3001" to-port="1" />
-		<edge from-layer="3001" from-port="2" to-layer="3209" to-port="0" />
-		<edge from-layer="3002" from-port="0" to-layer="3003" to-port="0" />
-		<edge from-layer="3003" from-port="1" to-layer="3004" to-port="1" />
-		<edge from-layer="3004" from-port="2" to-layer="3007" to-port="0" />
-		<edge from-layer="3005" from-port="0" to-layer="3006" to-port="0" />
-		<edge from-layer="3006" from-port="1" to-layer="3007" to-port="1" />
-		<edge from-layer="3007" from-port="2" to-layer="3016" to-port="0" />
+		<edge from-layer="2961" from-port="2" to-layer="2962" to-port="1" />
+		<edge from-layer="2962" from-port="2" to-layer="2971" to-port="2" />
+		<edge from-layer="2962" from-port="2" to-layer="5140" to-port="0" />
+		<edge from-layer="2963" from-port="0" to-layer="2970" to-port="1" />
+		<edge from-layer="2964" from-port="1" to-layer="2967" to-port="0" />
+		<edge from-layer="2965" from-port="0" to-layer="2967" to-port="1" />
+		<edge from-layer="2966" from-port="0" to-layer="2967" to-port="2" />
+		<edge from-layer="2967" from-port="3" to-layer="2968" to-port="0" />
+		<edge from-layer="2968" from-port="2" to-layer="2970" to-port="2" />
+		<edge from-layer="2969" from-port="0" to-layer="2970" to-port="3" />
+		<edge from-layer="2970" from-port="5" to-layer="2971" to-port="3" />
+		<edge from-layer="2971" from-port="4" to-layer="2973" to-port="0" />
+		<edge from-layer="2972" from-port="0" to-layer="2973" to-port="1" />
+		<edge from-layer="2973" from-port="2" to-layer="2975" to-port="0" />
+		<edge from-layer="2974" from-port="0" to-layer="2975" to-port="1" />
+		<edge from-layer="2975" from-port="2" to-layer="2984" to-port="0" />
+		<edge from-layer="2976" from-port="0" to-layer="2977" to-port="0" />
+		<edge from-layer="2977" from-port="1" to-layer="2980" to-port="0" />
+		<edge from-layer="2978" from-port="0" to-layer="2979" to-port="0" />
+		<edge from-layer="2979" from-port="1" to-layer="2980" to-port="1" />
+		<edge from-layer="2980" from-port="2" to-layer="2982" to-port="0" />
+		<edge from-layer="2981" from-port="0" to-layer="2982" to-port="1" />
+		<edge from-layer="2982" from-port="2" to-layer="2983" to-port="0" />
+		<edge from-layer="2983" from-port="1" to-layer="2984" to-port="1" />
+		<edge from-layer="2984" from-port="2" to-layer="2986" to-port="0" />
+		<edge from-layer="2985" from-port="0" to-layer="2986" to-port="1" />
+		<edge from-layer="2986" from-port="2" to-layer="2987" to-port="1" />
+		<edge from-layer="2987" from-port="2" to-layer="2988" to-port="0" />
+		<edge from-layer="2988" from-port="2" to-layer="2990" to-port="0" />
+		<edge from-layer="2988" from-port="2" to-layer="3146" to-port="1" />
+		<edge from-layer="2989" from-port="0" to-layer="2990" to-port="1" />
+		<edge from-layer="2990" from-port="2" to-layer="2992" to-port="0" />
+		<edge from-layer="2990" from-port="2" to-layer="3021" to-port="0" />
+		<edge from-layer="2991" from-port="0" to-layer="2992" to-port="1" />
+		<edge from-layer="2992" from-port="2" to-layer="2994" to-port="0" />
+		<edge from-layer="2993" from-port="0" to-layer="2994" to-port="1" />
+		<edge from-layer="2994" from-port="2" to-layer="3003" to-port="0" />
+		<edge from-layer="2995" from-port="0" to-layer="2996" to-port="0" />
+		<edge from-layer="2996" from-port="1" to-layer="2999" to-port="0" />
+		<edge from-layer="2997" from-port="0" to-layer="2998" to-port="0" />
+		<edge from-layer="2998" from-port="1" to-layer="2999" to-port="1" />
+		<edge from-layer="2999" from-port="2" to-layer="3001" to-port="0" />
+		<edge from-layer="3000" from-port="0" to-layer="3001" to-port="1" />
+		<edge from-layer="3001" from-port="2" to-layer="3002" to-port="0" />
+		<edge from-layer="3002" from-port="1" to-layer="3003" to-port="1" />
+		<edge from-layer="3003" from-port="2" to-layer="3006" to-port="0" />
+		<edge from-layer="3004" from-port="0" to-layer="3005" to-port="0" />
+		<edge from-layer="3005" from-port="1" to-layer="3006" to-port="1" />
+		<edge from-layer="3006" from-port="2" to-layer="3007" to-port="0" />
+		<edge from-layer="3007" from-port="1" to-layer="3016" to-port="0" />
 		<edge from-layer="3008" from-port="0" to-layer="3009" to-port="0" />
 		<edge from-layer="3009" from-port="1" to-layer="3012" to-port="0" />
 		<edge from-layer="3010" from-port="0" to-layer="3011" to-port="0" />
@@ -124907,2112 +84451,2030 @@
 		<edge from-layer="3016" from-port="2" to-layer="3019" to-port="0" />
 		<edge from-layer="3017" from-port="0" to-layer="3018" to-port="0" />
 		<edge from-layer="3018" from-port="1" to-layer="3019" to-port="1" />
-		<edge from-layer="3019" from-port="2" to-layer="3021" to-port="0" />
+		<edge from-layer="3019" from-port="2" to-layer="3145" to-port="0" />
 		<edge from-layer="3020" from-port="0" to-layer="3021" to-port="1" />
-		<edge from-layer="3021" from-port="2" to-layer="3025" to-port="0" />
-		<edge from-layer="3021" from-port="2" to-layer="3149" to-port="0" />
-		<edge from-layer="3021" from-port="2" to-layer="3038" to-port="0" />
-		<edge from-layer="3022" from-port="0" to-layer="3025" to-port="1" />
-		<edge from-layer="3023" from-port="0" to-layer="3025" to-port="2" />
-		<edge from-layer="3024" from-port="0" to-layer="3025" to-port="3" />
-		<edge from-layer="3025" from-port="4" to-layer="3027" to-port="0" />
-		<edge from-layer="3026" from-port="0" to-layer="3027" to-port="1" />
-		<edge from-layer="3027" from-port="2" to-layer="3031" to-port="0" />
-		<edge from-layer="3027" from-port="2" to-layer="3100" to-port="0" />
-		<edge from-layer="3028" from-port="0" to-layer="3031" to-port="1" />
-		<edge from-layer="3029" from-port="0" to-layer="3031" to-port="2" />
-		<edge from-layer="3030" from-port="0" to-layer="3031" to-port="3" />
-		<edge from-layer="3031" from-port="4" to-layer="3085" to-port="0" />
-		<edge from-layer="3031" from-port="4" to-layer="3079" to-port="0" />
-		<edge from-layer="3031" from-port="4" to-layer="3068" to-port="0" />
-		<edge from-layer="3031" from-port="4" to-layer="3065" to-port="0" />
-		<edge from-layer="3032" from-port="0" to-layer="3033" to-port="0" />
-		<edge from-layer="3033" from-port="1" to-layer="3061" to-port="0" />
-		<edge from-layer="3034" from-port="0" to-layer="3061" to-port="1" />
-		<edge from-layer="3035" from-port="0" to-layer="3038" to-port="1" />
-		<edge from-layer="3036" from-port="0" to-layer="3038" to-port="2" />
-		<edge from-layer="3037" from-port="0" to-layer="3038" to-port="3" />
-		<edge from-layer="3038" from-port="4" to-layer="3040" to-port="0" />
-		<edge from-layer="3039" from-port="0" to-layer="3040" to-port="1" />
-		<edge from-layer="3040" from-port="2" to-layer="3133" to-port="0" />
-		<edge from-layer="3040" from-port="2" to-layer="3041" to-port="0" />
-		<edge from-layer="3040" from-port="2" to-layer="3105" to-port="0" />
-		<edge from-layer="3041" from-port="1" to-layer="3043" to-port="0" />
-		<edge from-layer="3042" from-port="0" to-layer="3043" to-port="2" />
-		<edge from-layer="3043" from-port="3" to-layer="3057" to-port="0" />
-		<edge from-layer="3044" from-port="0" to-layer="3045" to-port="0" />
-		<edge from-layer="3045" from-port="1" to-layer="3050" to-port="0" />
-		<edge from-layer="3046" from-port="0" to-layer="3049" to-port="1" />
-		<edge from-layer="3047" from-port="0" to-layer="3049" to-port="2" />
-		<edge from-layer="3048" from-port="0" to-layer="3049" to-port="3" />
-		<edge from-layer="3049" from-port="4" to-layer="3050" to-port="1" />
-		<edge from-layer="3050" from-port="2" to-layer="3051" to-port="0" />
-		<edge from-layer="3051" from-port="1" to-layer="3053" to-port="0" />
-		<edge from-layer="3052" from-port="0" to-layer="3053" to-port="2" />
-		<edge from-layer="3053" from-port="3" to-layer="3054" to-port="0" />
-		<edge from-layer="3053" from-port="3" to-layer="3135" to-port="0" />
-		<edge from-layer="3054" from-port="1" to-layer="3056" to-port="0" />
-		<edge from-layer="3055" from-port="0" to-layer="3056" to-port="2" />
-		<edge from-layer="3056" from-port="3" to-layer="3057" to-port="1" />
-		<edge from-layer="3057" from-port="2" to-layer="3059" to-port="0" />
-		<edge from-layer="3058" from-port="0" to-layer="3059" to-port="1" />
-		<edge from-layer="3059" from-port="2" to-layer="3091" to-port="2" />
-		<edge from-layer="3059" from-port="2" to-layer="3061" to-port="2" />
-		<edge from-layer="3060" from-port="0" to-layer="3061" to-port="3" />
-		<edge from-layer="3061" from-port="4" to-layer="3063" to-port="0" />
-		<edge from-layer="3062" from-port="0" to-layer="3063" to-port="2" />
-		<edge from-layer="3063" from-port="3" to-layer="3064" to-port="0" />
-		<edge from-layer="3064" from-port="2" to-layer="3106" to-port="1" />
-		<edge from-layer="3064" from-port="2" to-layer="3065" to-port="1" />
-		<edge from-layer="3065" from-port="2" to-layer="3096" to-port="0" />
-		<edge from-layer="3066" from-port="0" to-layer="3076" to-port="0" />
-		<edge from-layer="3067" from-port="0" to-layer="3076" to-port="1" />
-		<edge from-layer="3068" from-port="1" to-layer="3070" to-port="0" />
-		<edge from-layer="3069" from-port="0" to-layer="3070" to-port="2" />
-		<edge from-layer="3070" from-port="3" to-layer="3071" to-port="0" />
+		<edge from-layer="3021" from-port="2" to-layer="3023" to-port="0" />
+		<edge from-layer="3022" from-port="0" to-layer="3023" to-port="1" />
+		<edge from-layer="3023" from-port="2" to-layer="3032" to-port="0" />
+		<edge from-layer="3024" from-port="0" to-layer="3025" to-port="0" />
+		<edge from-layer="3025" from-port="1" to-layer="3028" to-port="0" />
+		<edge from-layer="3026" from-port="0" to-layer="3027" to-port="0" />
+		<edge from-layer="3027" from-port="1" to-layer="3028" to-port="1" />
+		<edge from-layer="3028" from-port="2" to-layer="3030" to-port="0" />
+		<edge from-layer="3029" from-port="0" to-layer="3030" to-port="1" />
+		<edge from-layer="3030" from-port="2" to-layer="3031" to-port="0" />
+		<edge from-layer="3031" from-port="1" to-layer="3032" to-port="1" />
+		<edge from-layer="3032" from-port="2" to-layer="3034" to-port="0" />
+		<edge from-layer="3033" from-port="0" to-layer="3034" to-port="1" />
+		<edge from-layer="3034" from-port="2" to-layer="3036" to-port="0" />
+		<edge from-layer="3035" from-port="0" to-layer="3036" to-port="1" />
+		<edge from-layer="3036" from-port="2" to-layer="3078" to-port="0" />
+		<edge from-layer="3036" from-port="2" to-layer="3117" to-port="0" />
+		<edge from-layer="3036" from-port="2" to-layer="3040" to-port="0" />
+		<edge from-layer="3037" from-port="0" to-layer="3040" to-port="1" />
+		<edge from-layer="3038" from-port="0" to-layer="3040" to-port="2" />
+		<edge from-layer="3039" from-port="0" to-layer="3040" to-port="3" />
+		<edge from-layer="3040" from-port="5" to-layer="3042" to-port="0" />
+		<edge from-layer="3041" from-port="0" to-layer="3042" to-port="1" />
+		<edge from-layer="3042" from-port="2" to-layer="3046" to-port="0" />
+		<edge from-layer="3042" from-port="2" to-layer="3064" to-port="0" />
+		<edge from-layer="3043" from-port="0" to-layer="3046" to-port="1" />
+		<edge from-layer="3044" from-port="0" to-layer="3046" to-port="2" />
+		<edge from-layer="3045" from-port="0" to-layer="3046" to-port="3" />
+		<edge from-layer="3046" from-port="5" to-layer="3047" to-port="0" />
+		<edge from-layer="3046" from-port="5" to-layer="3051" to-port="0" />
+		<edge from-layer="3046" from-port="5" to-layer="3057" to-port="0" />
+		<edge from-layer="3047" from-port="2" to-layer="3060" to-port="0" />
+		<edge from-layer="3048" from-port="0" to-layer="3051" to-port="1" />
+		<edge from-layer="3049" from-port="0" to-layer="3051" to-port="2" />
+		<edge from-layer="3050" from-port="0" to-layer="3051" to-port="3" />
+		<edge from-layer="3051" from-port="5" to-layer="3053" to-port="0" />
+		<edge from-layer="3052" from-port="0" to-layer="3053" to-port="1" />
+		<edge from-layer="3053" from-port="2" to-layer="3058" to-port="0" />
+		<edge from-layer="3054" from-port="0" to-layer="3057" to-port="1" />
+		<edge from-layer="3055" from-port="0" to-layer="3057" to-port="2" />
+		<edge from-layer="3056" from-port="0" to-layer="3057" to-port="3" />
+		<edge from-layer="3057" from-port="5" to-layer="3058" to-port="1" />
+		<edge from-layer="3058" from-port="2" to-layer="3059" to-port="0" />
+		<edge from-layer="3059" from-port="2" to-layer="3060" to-port="1" />
+		<edge from-layer="3060" from-port="2" to-layer="3065" to-port="0" />
+		<edge from-layer="3061" from-port="0" to-layer="3064" to-port="1" />
+		<edge from-layer="3062" from-port="0" to-layer="3064" to-port="2" />
+		<edge from-layer="3063" from-port="0" to-layer="3064" to-port="3" />
+		<edge from-layer="3064" from-port="5" to-layer="3065" to-port="1" />
+		<edge from-layer="3065" from-port="2" to-layer="3129" to-port="0" />
+		<edge from-layer="3066" from-port="0" to-layer="3071" to-port="0" />
+		<edge from-layer="3067" from-port="0" to-layer="3070" to-port="1" />
+		<edge from-layer="3068" from-port="0" to-layer="3070" to-port="2" />
+		<edge from-layer="3069" from-port="0" to-layer="3070" to-port="3" />
+		<edge from-layer="3070" from-port="4" to-layer="3071" to-port="1" />
 		<edge from-layer="3071" from-port="2" to-layer="3072" to-port="0" />
 		<edge from-layer="3072" from-port="1" to-layer="3074" to-port="0" />
-		<edge from-layer="3073" from-port="0" to-layer="3074" to-port="1" />
-		<edge from-layer="3074" from-port="2" to-layer="3076" to-port="2" />
-		<edge from-layer="3075" from-port="0" to-layer="3076" to-port="3" />
-		<edge from-layer="3076" from-port="4" to-layer="3079" to-port="1" />
-		<edge from-layer="3076" from-port="4" to-layer="3085" to-port="2" />
-		<edge from-layer="3077" from-port="0" to-layer="3079" to-port="2" />
-		<edge from-layer="3078" from-port="0" to-layer="3079" to-port="3" />
-		<edge from-layer="3079" from-port="4" to-layer="3082" to-port="0" />
-		<edge from-layer="3080" from-port="0" to-layer="3081" to-port="0" />
-		<edge from-layer="3081" from-port="1" to-layer="3082" to-port="1" />
-		<edge from-layer="3082" from-port="2" to-layer="3086" to-port="0" />
-		<edge from-layer="3083" from-port="0" to-layer="3085" to-port="1" />
-		<edge from-layer="3084" from-port="0" to-layer="3085" to-port="3" />
-		<edge from-layer="3085" from-port="4" to-layer="3086" to-port="1" />
-		<edge from-layer="3086" from-port="2" to-layer="3095" to-port="0" />
-		<edge from-layer="3087" from-port="0" to-layer="3088" to-port="0" />
-		<edge from-layer="3088" from-port="1" to-layer="3091" to-port="0" />
-		<edge from-layer="3089" from-port="0" to-layer="3091" to-port="1" />
-		<edge from-layer="3090" from-port="0" to-layer="3091" to-port="3" />
-		<edge from-layer="3091" from-port="4" to-layer="3093" to-port="0" />
-		<edge from-layer="3092" from-port="0" to-layer="3093" to-port="2" />
-		<edge from-layer="3093" from-port="3" to-layer="3094" to-port="0" />
-		<edge from-layer="3094" from-port="2" to-layer="3095" to-port="1" />
-		<edge from-layer="3094" from-port="2" to-layer="3128" to-port="1" />
-		<edge from-layer="3095" from-port="2" to-layer="3096" to-port="1" />
-		<edge from-layer="3096" from-port="2" to-layer="3101" to-port="0" />
-		<edge from-layer="3097" from-port="0" to-layer="3100" to-port="1" />
-		<edge from-layer="3098" from-port="0" to-layer="3100" to-port="2" />
-		<edge from-layer="3099" from-port="0" to-layer="3100" to-port="3" />
-		<edge from-layer="3100" from-port="4" to-layer="3101" to-port="1" />
-		<edge from-layer="3101" from-port="2" to-layer="3192" to-port="0" />
-		<edge from-layer="3101" from-port="2" to-layer="3159" to-port="0" />
-		<edge from-layer="3102" from-port="0" to-layer="3105" to-port="1" />
-		<edge from-layer="3103" from-port="0" to-layer="3105" to-port="2" />
-		<edge from-layer="3104" from-port="0" to-layer="3105" to-port="3" />
-		<edge from-layer="3105" from-port="4" to-layer="3126" to-port="0" />
-		<edge from-layer="3105" from-port="4" to-layer="3120" to-port="0" />
-		<edge from-layer="3105" from-port="4" to-layer="3106" to-port="0" />
-		<edge from-layer="3105" from-port="4" to-layer="3109" to-port="0" />
-		<edge from-layer="3106" from-port="2" to-layer="3129" to-port="0" />
-		<edge from-layer="3107" from-port="0" to-layer="3117" to-port="0" />
-		<edge from-layer="3108" from-port="0" to-layer="3117" to-port="1" />
-		<edge from-layer="3109" from-port="1" to-layer="3111" to-port="0" />
-		<edge from-layer="3110" from-port="0" to-layer="3111" to-port="2" />
-		<edge from-layer="3111" from-port="3" to-layer="3112" to-port="0" />
-		<edge from-layer="3112" from-port="2" to-layer="3113" to-port="0" />
-		<edge from-layer="3113" from-port="1" to-layer="3115" to-port="0" />
-		<edge from-layer="3114" from-port="0" to-layer="3115" to-port="1" />
-		<edge from-layer="3115" from-port="2" to-layer="3117" to-port="2" />
+		<edge from-layer="3073" from-port="0" to-layer="3074" to-port="2" />
+		<edge from-layer="3074" from-port="3" to-layer="3122" to-port="0" />
+		<edge from-layer="3074" from-port="3" to-layer="3104" to-port="0" />
+		<edge from-layer="3075" from-port="0" to-layer="3078" to-port="1" />
+		<edge from-layer="3076" from-port="0" to-layer="3078" to-port="2" />
+		<edge from-layer="3077" from-port="0" to-layer="3078" to-port="3" />
+		<edge from-layer="3078" from-port="5" to-layer="3080" to-port="0" />
+		<edge from-layer="3079" from-port="0" to-layer="3080" to-port="1" />
+		<edge from-layer="3080" from-port="2" to-layer="3102" to-port="0" />
+		<edge from-layer="3080" from-port="2" to-layer="3084" to-port="0" />
+		<edge from-layer="3081" from-port="0" to-layer="3084" to-port="1" />
+		<edge from-layer="3082" from-port="0" to-layer="3084" to-port="2" />
+		<edge from-layer="3083" from-port="0" to-layer="3084" to-port="3" />
+		<edge from-layer="3084" from-port="5" to-layer="3085" to-port="0" />
+		<edge from-layer="3084" from-port="5" to-layer="3095" to-port="0" />
+		<edge from-layer="3084" from-port="5" to-layer="3089" to-port="0" />
+		<edge from-layer="3085" from-port="2" to-layer="3098" to-port="0" />
+		<edge from-layer="3086" from-port="0" to-layer="3089" to-port="1" />
+		<edge from-layer="3087" from-port="0" to-layer="3089" to-port="2" />
+		<edge from-layer="3088" from-port="0" to-layer="3089" to-port="3" />
+		<edge from-layer="3089" from-port="5" to-layer="3091" to-port="0" />
+		<edge from-layer="3090" from-port="0" to-layer="3091" to-port="1" />
+		<edge from-layer="3091" from-port="2" to-layer="3096" to-port="0" />
+		<edge from-layer="3092" from-port="0" to-layer="3095" to-port="1" />
+		<edge from-layer="3093" from-port="0" to-layer="3095" to-port="2" />
+		<edge from-layer="3094" from-port="0" to-layer="3095" to-port="3" />
+		<edge from-layer="3095" from-port="5" to-layer="3096" to-port="1" />
+		<edge from-layer="3096" from-port="2" to-layer="3097" to-port="0" />
+		<edge from-layer="3097" from-port="2" to-layer="3098" to-port="1" />
+		<edge from-layer="3098" from-port="2" to-layer="3103" to-port="0" />
+		<edge from-layer="3099" from-port="0" to-layer="3102" to-port="1" />
+		<edge from-layer="3100" from-port="0" to-layer="3102" to-port="2" />
+		<edge from-layer="3101" from-port="0" to-layer="3102" to-port="3" />
+		<edge from-layer="3102" from-port="5" to-layer="3103" to-port="1" />
+		<edge from-layer="3103" from-port="2" to-layer="3104" to-port="1" />
+		<edge from-layer="3104" from-port="2" to-layer="5139" to-port="0" />
+		<edge from-layer="3104" from-port="2" to-layer="3129" to-port="1" />
+		<edge from-layer="3105" from-port="0" to-layer="3110" to-port="0" />
+		<edge from-layer="3106" from-port="0" to-layer="3109" to-port="1" />
+		<edge from-layer="3107" from-port="0" to-layer="3109" to-port="2" />
+		<edge from-layer="3108" from-port="0" to-layer="3109" to-port="3" />
+		<edge from-layer="3109" from-port="4" to-layer="3110" to-port="1" />
+		<edge from-layer="3110" from-port="2" to-layer="3111" to-port="0" />
+		<edge from-layer="3111" from-port="1" to-layer="3113" to-port="0" />
+		<edge from-layer="3112" from-port="0" to-layer="3113" to-port="2" />
+		<edge from-layer="3113" from-port="3" to-layer="3120" to-port="0" />
+		<edge from-layer="3114" from-port="0" to-layer="3117" to-port="1" />
+		<edge from-layer="3115" from-port="0" to-layer="3117" to-port="2" />
 		<edge from-layer="3116" from-port="0" to-layer="3117" to-port="3" />
-		<edge from-layer="3117" from-port="4" to-layer="3120" to-port="1" />
-		<edge from-layer="3117" from-port="4" to-layer="3126" to-port="2" />
-		<edge from-layer="3118" from-port="0" to-layer="3120" to-port="2" />
-		<edge from-layer="3119" from-port="0" to-layer="3120" to-port="3" />
-		<edge from-layer="3120" from-port="4" to-layer="3123" to-port="0" />
-		<edge from-layer="3121" from-port="0" to-layer="3122" to-port="0" />
-		<edge from-layer="3122" from-port="1" to-layer="3123" to-port="1" />
-		<edge from-layer="3123" from-port="2" to-layer="3127" to-port="0" />
-		<edge from-layer="3124" from-port="0" to-layer="3126" to-port="1" />
-		<edge from-layer="3125" from-port="0" to-layer="3126" to-port="3" />
-		<edge from-layer="3126" from-port="4" to-layer="3127" to-port="1" />
-		<edge from-layer="3127" from-port="2" to-layer="3128" to-port="0" />
-		<edge from-layer="3128" from-port="2" to-layer="3129" to-port="1" />
-		<edge from-layer="3129" from-port="2" to-layer="3134" to-port="0" />
-		<edge from-layer="3130" from-port="0" to-layer="3133" to-port="1" />
-		<edge from-layer="3131" from-port="0" to-layer="3133" to-port="2" />
-		<edge from-layer="3132" from-port="0" to-layer="3133" to-port="3" />
-		<edge from-layer="3133" from-port="4" to-layer="3134" to-port="1" />
-		<edge from-layer="3134" from-port="2" to-layer="3135" to-port="1" />
-		<edge from-layer="3135" from-port="2" to-layer="3156" to-port="0" />
-		<edge from-layer="3135" from-port="2" to-layer="3192" to-port="1" />
-		<edge from-layer="3135" from-port="2" to-layer="7714" to-port="0" />
+		<edge from-layer="3117" from-port="5" to-layer="3119" to-port="0" />
+		<edge from-layer="3118" from-port="0" to-layer="3119" to-port="1" />
+		<edge from-layer="3119" from-port="2" to-layer="3120" to-port="1" />
+		<edge from-layer="3120" from-port="2" to-layer="3129" to-port="2" />
+		<edge from-layer="3120" from-port="2" to-layer="5138" to-port="0" />
+		<edge from-layer="3121" from-port="0" to-layer="3128" to-port="1" />
+		<edge from-layer="3122" from-port="1" to-layer="3125" to-port="0" />
+		<edge from-layer="3123" from-port="0" to-layer="3125" to-port="1" />
+		<edge from-layer="3124" from-port="0" to-layer="3125" to-port="2" />
+		<edge from-layer="3125" from-port="3" to-layer="3126" to-port="0" />
+		<edge from-layer="3126" from-port="2" to-layer="3128" to-port="2" />
+		<edge from-layer="3127" from-port="0" to-layer="3128" to-port="3" />
+		<edge from-layer="3128" from-port="5" to-layer="3129" to-port="3" />
+		<edge from-layer="3129" from-port="4" to-layer="3131" to-port="0" />
+		<edge from-layer="3130" from-port="0" to-layer="3131" to-port="1" />
+		<edge from-layer="3131" from-port="2" to-layer="3133" to-port="0" />
+		<edge from-layer="3132" from-port="0" to-layer="3133" to-port="1" />
+		<edge from-layer="3133" from-port="2" to-layer="3142" to-port="0" />
+		<edge from-layer="3134" from-port="0" to-layer="3135" to-port="0" />
+		<edge from-layer="3135" from-port="1" to-layer="3138" to-port="0" />
 		<edge from-layer="3136" from-port="0" to-layer="3137" to-port="0" />
-		<edge from-layer="3137" from-port="1" to-layer="3142" to-port="0" />
-		<edge from-layer="3138" from-port="0" to-layer="3141" to-port="1" />
-		<edge from-layer="3139" from-port="0" to-layer="3141" to-port="2" />
-		<edge from-layer="3140" from-port="0" to-layer="3141" to-port="3" />
-		<edge from-layer="3141" from-port="4" to-layer="3142" to-port="1" />
-		<edge from-layer="3142" from-port="2" to-layer="3143" to-port="0" />
-		<edge from-layer="3143" from-port="1" to-layer="3145" to-port="0" />
-		<edge from-layer="3144" from-port="0" to-layer="3145" to-port="2" />
-		<edge from-layer="3145" from-port="3" to-layer="3152" to-port="0" />
-		<edge from-layer="3146" from-port="0" to-layer="3149" to-port="1" />
-		<edge from-layer="3147" from-port="0" to-layer="3149" to-port="2" />
-		<edge from-layer="3148" from-port="0" to-layer="3149" to-port="3" />
-		<edge from-layer="3149" from-port="4" to-layer="3151" to-port="0" />
-		<edge from-layer="3150" from-port="0" to-layer="3151" to-port="1" />
-		<edge from-layer="3151" from-port="2" to-layer="3152" to-port="1" />
-		<edge from-layer="3152" from-port="2" to-layer="3192" to-port="2" />
-		<edge from-layer="3152" from-port="2" to-layer="7713" to-port="0" />
-		<edge from-layer="3153" from-port="0" to-layer="3171" to-port="0" />
-		<edge from-layer="3154" from-port="0" to-layer="3166" to-port="0" />
-		<edge from-layer="3155" from-port="0" to-layer="3169" to-port="1" />
-		<edge from-layer="3155" from-port="0" to-layer="3166" to-port="1" />
-		<edge from-layer="3156" from-port="1" to-layer="3158" to-port="0" />
-		<edge from-layer="3157" from-port="0" to-layer="3158" to-port="2" />
-		<edge from-layer="3158" from-port="3" to-layer="3162" to-port="0" />
-		<edge from-layer="3158" from-port="3" to-layer="3168" to-port="0" />
-		<edge from-layer="3159" from-port="1" to-layer="3187" to-port="0" />
-		<edge from-layer="3159" from-port="1" to-layer="3161" to-port="0" />
-		<edge from-layer="3160" from-port="0" to-layer="3161" to-port="2" />
-		<edge from-layer="3161" from-port="3" to-layer="3162" to-port="1" />
-		<edge from-layer="3162" from-port="2" to-layer="3164" to-port="0" />
-		<edge from-layer="3163" from-port="0" to-layer="3168" to-port="1" />
-		<edge from-layer="3163" from-port="0" to-layer="3164" to-port="1" />
-		<edge from-layer="3164" from-port="2" to-layer="3166" to-port="2" />
-		<edge from-layer="3165" from-port="0" to-layer="3169" to-port="3" />
-		<edge from-layer="3165" from-port="0" to-layer="3166" to-port="3" />
-		<edge from-layer="3166" from-port="4" to-layer="3171" to-port="1" />
-		<edge from-layer="3167" from-port="0" to-layer="3169" to-port="0" />
-		<edge from-layer="3168" from-port="2" to-layer="3169" to-port="2" />
-		<edge from-layer="3168" from-port="2" to-layer="3176" to-port="2" />
-		<edge from-layer="3169" from-port="4" to-layer="3171" to-port="2" />
-		<edge from-layer="3170" from-port="0" to-layer="3171" to-port="3" />
-		<edge from-layer="3171" from-port="4" to-layer="3178" to-port="0" />
-		<edge from-layer="3172" from-port="0" to-layer="3178" to-port="1" />
-		<edge from-layer="3173" from-port="0" to-layer="3176" to-port="0" />
-		<edge from-layer="3174" from-port="0" to-layer="3176" to-port="1" />
-		<edge from-layer="3175" from-port="0" to-layer="3176" to-port="3" />
-		<edge from-layer="3176" from-port="4" to-layer="3178" to-port="2" />
-		<edge from-layer="3177" from-port="0" to-layer="3178" to-port="3" />
-		<edge from-layer="3178" from-port="4" to-layer="3182" to-port="0" />
-		<edge from-layer="3179" from-port="0" to-layer="3180" to-port="0" />
-		<edge from-layer="3180" from-port="1" to-layer="3182" to-port="1" />
-		<edge from-layer="3181" from-port="0" to-layer="3182" to-port="2" />
-		<edge from-layer="3182" from-port="3" to-layer="3183" to-port="0" />
-		<edge from-layer="3183" from-port="2" to-layer="3184" to-port="0" />
-		<edge from-layer="3183" from-port="2" to-layer="3191" to-port="0" />
-		<edge from-layer="3184" from-port="1" to-layer="3190" to-port="0" />
-		<edge from-layer="3185" from-port="0" to-layer="3187" to-port="1" />
-		<edge from-layer="3186" from-port="0" to-layer="3187" to-port="2" />
-		<edge from-layer="3187" from-port="3" to-layer="3188" to-port="0" />
-		<edge from-layer="3188" from-port="4" to-layer="3189" to-port="0" />
+		<edge from-layer="3137" from-port="1" to-layer="3138" to-port="1" />
+		<edge from-layer="3138" from-port="2" to-layer="3140" to-port="0" />
+		<edge from-layer="3139" from-port="0" to-layer="3140" to-port="1" />
+		<edge from-layer="3140" from-port="2" to-layer="3141" to-port="0" />
+		<edge from-layer="3141" from-port="1" to-layer="3142" to-port="1" />
+		<edge from-layer="3142" from-port="2" to-layer="3144" to-port="0" />
+		<edge from-layer="3143" from-port="0" to-layer="3144" to-port="1" />
+		<edge from-layer="3144" from-port="2" to-layer="3145" to-port="1" />
+		<edge from-layer="3145" from-port="2" to-layer="3146" to-port="0" />
+		<edge from-layer="3146" from-port="2" to-layer="3148" to-port="0" />
+		<edge from-layer="3146" from-port="2" to-layer="3304" to-port="1" />
+		<edge from-layer="3147" from-port="0" to-layer="3148" to-port="1" />
+		<edge from-layer="3148" from-port="2" to-layer="3179" to-port="0" />
+		<edge from-layer="3148" from-port="2" to-layer="3150" to-port="0" />
+		<edge from-layer="3149" from-port="0" to-layer="3150" to-port="1" />
+		<edge from-layer="3150" from-port="2" to-layer="3152" to-port="0" />
+		<edge from-layer="3151" from-port="0" to-layer="3152" to-port="1" />
+		<edge from-layer="3152" from-port="2" to-layer="3161" to-port="0" />
+		<edge from-layer="3153" from-port="0" to-layer="3154" to-port="0" />
+		<edge from-layer="3154" from-port="1" to-layer="3157" to-port="0" />
+		<edge from-layer="3155" from-port="0" to-layer="3156" to-port="0" />
+		<edge from-layer="3156" from-port="1" to-layer="3157" to-port="1" />
+		<edge from-layer="3157" from-port="2" to-layer="3159" to-port="0" />
+		<edge from-layer="3158" from-port="0" to-layer="3159" to-port="1" />
+		<edge from-layer="3159" from-port="2" to-layer="3160" to-port="0" />
+		<edge from-layer="3160" from-port="1" to-layer="3161" to-port="1" />
+		<edge from-layer="3161" from-port="2" to-layer="3164" to-port="0" />
+		<edge from-layer="3162" from-port="0" to-layer="3163" to-port="0" />
+		<edge from-layer="3163" from-port="1" to-layer="3164" to-port="1" />
+		<edge from-layer="3164" from-port="2" to-layer="3165" to-port="0" />
+		<edge from-layer="3165" from-port="1" to-layer="3174" to-port="0" />
+		<edge from-layer="3166" from-port="0" to-layer="3167" to-port="0" />
+		<edge from-layer="3167" from-port="1" to-layer="3170" to-port="0" />
+		<edge from-layer="3168" from-port="0" to-layer="3169" to-port="0" />
+		<edge from-layer="3169" from-port="1" to-layer="3170" to-port="1" />
+		<edge from-layer="3170" from-port="2" to-layer="3172" to-port="0" />
+		<edge from-layer="3171" from-port="0" to-layer="3172" to-port="1" />
+		<edge from-layer="3172" from-port="2" to-layer="3173" to-port="0" />
+		<edge from-layer="3173" from-port="1" to-layer="3174" to-port="1" />
+		<edge from-layer="3174" from-port="2" to-layer="3177" to-port="0" />
+		<edge from-layer="3175" from-port="0" to-layer="3176" to-port="0" />
+		<edge from-layer="3176" from-port="1" to-layer="3177" to-port="1" />
+		<edge from-layer="3177" from-port="2" to-layer="3303" to-port="0" />
+		<edge from-layer="3178" from-port="0" to-layer="3179" to-port="1" />
+		<edge from-layer="3179" from-port="2" to-layer="3181" to-port="0" />
+		<edge from-layer="3180" from-port="0" to-layer="3181" to-port="1" />
+		<edge from-layer="3181" from-port="2" to-layer="3190" to-port="0" />
+		<edge from-layer="3182" from-port="0" to-layer="3183" to-port="0" />
+		<edge from-layer="3183" from-port="1" to-layer="3186" to-port="0" />
+		<edge from-layer="3184" from-port="0" to-layer="3185" to-port="0" />
+		<edge from-layer="3185" from-port="1" to-layer="3186" to-port="1" />
+		<edge from-layer="3186" from-port="2" to-layer="3188" to-port="0" />
+		<edge from-layer="3187" from-port="0" to-layer="3188" to-port="1" />
+		<edge from-layer="3188" from-port="2" to-layer="3189" to-port="0" />
 		<edge from-layer="3189" from-port="1" to-layer="3190" to-port="1" />
-		<edge from-layer="3190" from-port="2" to-layer="3191" to-port="1" />
-		<edge from-layer="3191" from-port="2" to-layer="3192" to-port="3" />
-		<edge from-layer="3192" from-port="4" to-layer="3194" to-port="0" />
+		<edge from-layer="3190" from-port="2" to-layer="3192" to-port="0" />
+		<edge from-layer="3191" from-port="0" to-layer="3192" to-port="1" />
+		<edge from-layer="3192" from-port="2" to-layer="3194" to-port="0" />
 		<edge from-layer="3193" from-port="0" to-layer="3194" to-port="1" />
-		<edge from-layer="3194" from-port="2" to-layer="3196" to-port="0" />
-		<edge from-layer="3195" from-port="0" to-layer="3196" to-port="1" />
-		<edge from-layer="3196" from-port="2" to-layer="3205" to-port="0" />
-		<edge from-layer="3197" from-port="0" to-layer="3198" to-port="0" />
-		<edge from-layer="3198" from-port="1" to-layer="3201" to-port="0" />
-		<edge from-layer="3199" from-port="0" to-layer="3200" to-port="0" />
-		<edge from-layer="3200" from-port="1" to-layer="3201" to-port="1" />
-		<edge from-layer="3201" from-port="2" to-layer="3203" to-port="0" />
-		<edge from-layer="3202" from-port="0" to-layer="3203" to-port="1" />
-		<edge from-layer="3203" from-port="2" to-layer="3204" to-port="0" />
-		<edge from-layer="3204" from-port="1" to-layer="3205" to-port="1" />
-		<edge from-layer="3205" from-port="2" to-layer="3208" to-port="0" />
-		<edge from-layer="3206" from-port="0" to-layer="3207" to-port="0" />
-		<edge from-layer="3207" from-port="1" to-layer="3208" to-port="1" />
-		<edge from-layer="3208" from-port="2" to-layer="3209" to-port="1" />
-		<edge from-layer="3209" from-port="2" to-layer="3210" to-port="0" />
-		<edge from-layer="3210" from-port="2" to-layer="3212" to-port="0" />
-		<edge from-layer="3210" from-port="2" to-layer="3452" to-port="1" />
-		<edge from-layer="3211" from-port="0" to-layer="3212" to-port="1" />
-		<edge from-layer="3212" from-port="2" to-layer="3246" to-port="0" />
-		<edge from-layer="3212" from-port="2" to-layer="3215" to-port="0" />
-		<edge from-layer="3213" from-port="0" to-layer="3214" to-port="0" />
-		<edge from-layer="3214" from-port="1" to-layer="3215" to-port="1" />
-		<edge from-layer="3215" from-port="2" to-layer="3218" to-port="0" />
-		<edge from-layer="3216" from-port="0" to-layer="3217" to-port="0" />
-		<edge from-layer="3217" from-port="1" to-layer="3218" to-port="1" />
-		<edge from-layer="3218" from-port="2" to-layer="3227" to-port="0" />
-		<edge from-layer="3219" from-port="0" to-layer="3220" to-port="0" />
-		<edge from-layer="3220" from-port="1" to-layer="3223" to-port="0" />
-		<edge from-layer="3221" from-port="0" to-layer="3222" to-port="0" />
-		<edge from-layer="3222" from-port="1" to-layer="3223" to-port="1" />
-		<edge from-layer="3223" from-port="2" to-layer="3225" to-port="0" />
-		<edge from-layer="3224" from-port="0" to-layer="3225" to-port="1" />
-		<edge from-layer="3225" from-port="2" to-layer="3226" to-port="0" />
-		<edge from-layer="3226" from-port="1" to-layer="3227" to-port="1" />
-		<edge from-layer="3227" from-port="2" to-layer="3230" to-port="0" />
-		<edge from-layer="3228" from-port="0" to-layer="3229" to-port="0" />
-		<edge from-layer="3229" from-port="1" to-layer="3230" to-port="1" />
-		<edge from-layer="3230" from-port="2" to-layer="3231" to-port="0" />
-		<edge from-layer="3231" from-port="1" to-layer="3240" to-port="0" />
-		<edge from-layer="3232" from-port="0" to-layer="3233" to-port="0" />
-		<edge from-layer="3233" from-port="1" to-layer="3236" to-port="0" />
-		<edge from-layer="3234" from-port="0" to-layer="3235" to-port="0" />
-		<edge from-layer="3235" from-port="1" to-layer="3236" to-port="1" />
-		<edge from-layer="3236" from-port="2" to-layer="3238" to-port="0" />
+		<edge from-layer="3194" from-port="2" to-layer="3275" to-port="0" />
+		<edge from-layer="3194" from-port="2" to-layer="3236" to-port="0" />
+		<edge from-layer="3194" from-port="2" to-layer="3198" to-port="0" />
+		<edge from-layer="3195" from-port="0" to-layer="3198" to-port="1" />
+		<edge from-layer="3196" from-port="0" to-layer="3198" to-port="2" />
+		<edge from-layer="3197" from-port="0" to-layer="3198" to-port="3" />
+		<edge from-layer="3198" from-port="5" to-layer="3200" to-port="0" />
+		<edge from-layer="3199" from-port="0" to-layer="3200" to-port="1" />
+		<edge from-layer="3200" from-port="2" to-layer="3204" to-port="0" />
+		<edge from-layer="3200" from-port="2" to-layer="3222" to-port="0" />
+		<edge from-layer="3201" from-port="0" to-layer="3204" to-port="1" />
+		<edge from-layer="3202" from-port="0" to-layer="3204" to-port="2" />
+		<edge from-layer="3203" from-port="0" to-layer="3204" to-port="3" />
+		<edge from-layer="3204" from-port="5" to-layer="3215" to-port="0" />
+		<edge from-layer="3204" from-port="5" to-layer="3209" to-port="0" />
+		<edge from-layer="3204" from-port="5" to-layer="3205" to-port="0" />
+		<edge from-layer="3205" from-port="2" to-layer="3218" to-port="0" />
+		<edge from-layer="3206" from-port="0" to-layer="3209" to-port="1" />
+		<edge from-layer="3207" from-port="0" to-layer="3209" to-port="2" />
+		<edge from-layer="3208" from-port="0" to-layer="3209" to-port="3" />
+		<edge from-layer="3209" from-port="5" to-layer="3211" to-port="0" />
+		<edge from-layer="3210" from-port="0" to-layer="3211" to-port="1" />
+		<edge from-layer="3211" from-port="2" to-layer="3216" to-port="0" />
+		<edge from-layer="3212" from-port="0" to-layer="3215" to-port="1" />
+		<edge from-layer="3213" from-port="0" to-layer="3215" to-port="2" />
+		<edge from-layer="3214" from-port="0" to-layer="3215" to-port="3" />
+		<edge from-layer="3215" from-port="5" to-layer="3216" to-port="1" />
+		<edge from-layer="3216" from-port="2" to-layer="3217" to-port="0" />
+		<edge from-layer="3217" from-port="2" to-layer="3218" to-port="1" />
+		<edge from-layer="3218" from-port="2" to-layer="3223" to-port="0" />
+		<edge from-layer="3219" from-port="0" to-layer="3222" to-port="1" />
+		<edge from-layer="3220" from-port="0" to-layer="3222" to-port="2" />
+		<edge from-layer="3221" from-port="0" to-layer="3222" to-port="3" />
+		<edge from-layer="3222" from-port="5" to-layer="3223" to-port="1" />
+		<edge from-layer="3223" from-port="2" to-layer="3287" to-port="0" />
+		<edge from-layer="3224" from-port="0" to-layer="3229" to-port="0" />
+		<edge from-layer="3225" from-port="0" to-layer="3228" to-port="1" />
+		<edge from-layer="3226" from-port="0" to-layer="3228" to-port="2" />
+		<edge from-layer="3227" from-port="0" to-layer="3228" to-port="3" />
+		<edge from-layer="3228" from-port="4" to-layer="3229" to-port="1" />
+		<edge from-layer="3229" from-port="2" to-layer="3230" to-port="0" />
+		<edge from-layer="3230" from-port="1" to-layer="3232" to-port="0" />
+		<edge from-layer="3231" from-port="0" to-layer="3232" to-port="2" />
+		<edge from-layer="3232" from-port="3" to-layer="3280" to-port="0" />
+		<edge from-layer="3232" from-port="3" to-layer="3262" to-port="0" />
+		<edge from-layer="3233" from-port="0" to-layer="3236" to-port="1" />
+		<edge from-layer="3234" from-port="0" to-layer="3236" to-port="2" />
+		<edge from-layer="3235" from-port="0" to-layer="3236" to-port="3" />
+		<edge from-layer="3236" from-port="5" to-layer="3238" to-port="0" />
 		<edge from-layer="3237" from-port="0" to-layer="3238" to-port="1" />
-		<edge from-layer="3238" from-port="2" to-layer="3239" to-port="0" />
-		<edge from-layer="3239" from-port="1" to-layer="3240" to-port="1" />
-		<edge from-layer="3240" from-port="2" to-layer="3243" to-port="0" />
-		<edge from-layer="3241" from-port="0" to-layer="3242" to-port="0" />
-		<edge from-layer="3242" from-port="1" to-layer="3243" to-port="1" />
-		<edge from-layer="3243" from-port="2" to-layer="3451" to-port="0" />
-		<edge from-layer="3244" from-port="0" to-layer="3245" to-port="0" />
-		<edge from-layer="3245" from-port="1" to-layer="3246" to-port="1" />
-		<edge from-layer="3246" from-port="2" to-layer="3249" to-port="0" />
-		<edge from-layer="3247" from-port="0" to-layer="3248" to-port="0" />
-		<edge from-layer="3248" from-port="1" to-layer="3249" to-port="1" />
-		<edge from-layer="3249" from-port="2" to-layer="3258" to-port="0" />
-		<edge from-layer="3250" from-port="0" to-layer="3251" to-port="0" />
-		<edge from-layer="3251" from-port="1" to-layer="3254" to-port="0" />
-		<edge from-layer="3252" from-port="0" to-layer="3253" to-port="0" />
-		<edge from-layer="3253" from-port="1" to-layer="3254" to-port="1" />
-		<edge from-layer="3254" from-port="2" to-layer="3256" to-port="0" />
-		<edge from-layer="3255" from-port="0" to-layer="3256" to-port="1" />
-		<edge from-layer="3256" from-port="2" to-layer="3257" to-port="0" />
-		<edge from-layer="3257" from-port="1" to-layer="3258" to-port="1" />
-		<edge from-layer="3258" from-port="2" to-layer="3261" to-port="0" />
-		<edge from-layer="3259" from-port="0" to-layer="3260" to-port="0" />
-		<edge from-layer="3260" from-port="1" to-layer="3261" to-port="1" />
-		<edge from-layer="3261" from-port="2" to-layer="3263" to-port="0" />
-		<edge from-layer="3262" from-port="0" to-layer="3263" to-port="1" />
-		<edge from-layer="3263" from-port="2" to-layer="3280" to-port="0" />
-		<edge from-layer="3263" from-port="2" to-layer="3391" to-port="0" />
-		<edge from-layer="3263" from-port="2" to-layer="3267" to-port="0" />
+		<edge from-layer="3238" from-port="2" to-layer="3242" to-port="0" />
+		<edge from-layer="3238" from-port="2" to-layer="3260" to-port="0" />
+		<edge from-layer="3239" from-port="0" to-layer="3242" to-port="1" />
+		<edge from-layer="3240" from-port="0" to-layer="3242" to-port="2" />
+		<edge from-layer="3241" from-port="0" to-layer="3242" to-port="3" />
+		<edge from-layer="3242" from-port="5" to-layer="3243" to-port="0" />
+		<edge from-layer="3242" from-port="5" to-layer="3253" to-port="0" />
+		<edge from-layer="3242" from-port="5" to-layer="3247" to-port="0" />
+		<edge from-layer="3243" from-port="2" to-layer="3256" to-port="0" />
+		<edge from-layer="3244" from-port="0" to-layer="3247" to-port="1" />
+		<edge from-layer="3245" from-port="0" to-layer="3247" to-port="2" />
+		<edge from-layer="3246" from-port="0" to-layer="3247" to-port="3" />
+		<edge from-layer="3247" from-port="5" to-layer="3249" to-port="0" />
+		<edge from-layer="3248" from-port="0" to-layer="3249" to-port="1" />
+		<edge from-layer="3249" from-port="2" to-layer="3254" to-port="0" />
+		<edge from-layer="3250" from-port="0" to-layer="3253" to-port="1" />
+		<edge from-layer="3251" from-port="0" to-layer="3253" to-port="2" />
+		<edge from-layer="3252" from-port="0" to-layer="3253" to-port="3" />
+		<edge from-layer="3253" from-port="5" to-layer="3254" to-port="1" />
+		<edge from-layer="3254" from-port="2" to-layer="3255" to-port="0" />
+		<edge from-layer="3255" from-port="2" to-layer="3256" to-port="1" />
+		<edge from-layer="3256" from-port="2" to-layer="3261" to-port="0" />
+		<edge from-layer="3257" from-port="0" to-layer="3260" to-port="1" />
+		<edge from-layer="3258" from-port="0" to-layer="3260" to-port="2" />
+		<edge from-layer="3259" from-port="0" to-layer="3260" to-port="3" />
+		<edge from-layer="3260" from-port="5" to-layer="3261" to-port="1" />
+		<edge from-layer="3261" from-port="2" to-layer="3262" to-port="1" />
+		<edge from-layer="3262" from-port="2" to-layer="5137" to-port="0" />
+		<edge from-layer="3262" from-port="2" to-layer="3287" to-port="1" />
+		<edge from-layer="3263" from-port="0" to-layer="3268" to-port="0" />
 		<edge from-layer="3264" from-port="0" to-layer="3267" to-port="1" />
 		<edge from-layer="3265" from-port="0" to-layer="3267" to-port="2" />
 		<edge from-layer="3266" from-port="0" to-layer="3267" to-port="3" />
-		<edge from-layer="3267" from-port="4" to-layer="3269" to-port="0" />
-		<edge from-layer="3268" from-port="0" to-layer="3269" to-port="1" />
-		<edge from-layer="3269" from-port="2" to-layer="3342" to-port="0" />
-		<edge from-layer="3269" from-port="2" to-layer="3273" to-port="0" />
-		<edge from-layer="3270" from-port="0" to-layer="3273" to-port="1" />
-		<edge from-layer="3271" from-port="0" to-layer="3273" to-port="2" />
-		<edge from-layer="3272" from-port="0" to-layer="3273" to-port="3" />
-		<edge from-layer="3273" from-port="4" to-layer="3307" to-port="0" />
-		<edge from-layer="3273" from-port="4" to-layer="3310" to-port="0" />
-		<edge from-layer="3273" from-port="4" to-layer="3327" to-port="0" />
-		<edge from-layer="3273" from-port="4" to-layer="3321" to-port="0" />
-		<edge from-layer="3274" from-port="0" to-layer="3275" to-port="0" />
-		<edge from-layer="3275" from-port="1" to-layer="3303" to-port="0" />
-		<edge from-layer="3276" from-port="0" to-layer="3303" to-port="1" />
-		<edge from-layer="3277" from-port="0" to-layer="3280" to-port="1" />
-		<edge from-layer="3278" from-port="0" to-layer="3280" to-port="2" />
-		<edge from-layer="3279" from-port="0" to-layer="3280" to-port="3" />
-		<edge from-layer="3280" from-port="4" to-layer="3282" to-port="0" />
-		<edge from-layer="3281" from-port="0" to-layer="3282" to-port="1" />
-		<edge from-layer="3282" from-port="2" to-layer="3375" to-port="0" />
-		<edge from-layer="3282" from-port="2" to-layer="3283" to-port="0" />
-		<edge from-layer="3282" from-port="2" to-layer="3347" to-port="0" />
-		<edge from-layer="3283" from-port="1" to-layer="3285" to-port="0" />
-		<edge from-layer="3284" from-port="0" to-layer="3285" to-port="2" />
-		<edge from-layer="3285" from-port="3" to-layer="3299" to-port="0" />
-		<edge from-layer="3286" from-port="0" to-layer="3287" to-port="0" />
-		<edge from-layer="3287" from-port="1" to-layer="3292" to-port="0" />
-		<edge from-layer="3288" from-port="0" to-layer="3291" to-port="1" />
-		<edge from-layer="3289" from-port="0" to-layer="3291" to-port="2" />
-		<edge from-layer="3290" from-port="0" to-layer="3291" to-port="3" />
-		<edge from-layer="3291" from-port="4" to-layer="3292" to-port="1" />
-		<edge from-layer="3292" from-port="2" to-layer="3293" to-port="0" />
-		<edge from-layer="3293" from-port="1" to-layer="3295" to-port="0" />
-		<edge from-layer="3294" from-port="0" to-layer="3295" to-port="2" />
-		<edge from-layer="3295" from-port="3" to-layer="3296" to-port="0" />
-		<edge from-layer="3295" from-port="3" to-layer="3377" to-port="0" />
-		<edge from-layer="3296" from-port="1" to-layer="3298" to-port="0" />
-		<edge from-layer="3297" from-port="0" to-layer="3298" to-port="2" />
-		<edge from-layer="3298" from-port="3" to-layer="3299" to-port="1" />
-		<edge from-layer="3299" from-port="2" to-layer="3301" to-port="0" />
-		<edge from-layer="3300" from-port="0" to-layer="3301" to-port="1" />
-		<edge from-layer="3301" from-port="2" to-layer="3333" to-port="2" />
-		<edge from-layer="3301" from-port="2" to-layer="3303" to-port="2" />
-		<edge from-layer="3302" from-port="0" to-layer="3303" to-port="3" />
-		<edge from-layer="3303" from-port="4" to-layer="3305" to-port="0" />
-		<edge from-layer="3304" from-port="0" to-layer="3305" to-port="2" />
-		<edge from-layer="3305" from-port="3" to-layer="3306" to-port="0" />
-		<edge from-layer="3306" from-port="2" to-layer="3348" to-port="1" />
-		<edge from-layer="3306" from-port="2" to-layer="3307" to-port="1" />
-		<edge from-layer="3307" from-port="2" to-layer="3338" to-port="0" />
-		<edge from-layer="3308" from-port="0" to-layer="3318" to-port="0" />
-		<edge from-layer="3309" from-port="0" to-layer="3318" to-port="1" />
-		<edge from-layer="3310" from-port="1" to-layer="3312" to-port="0" />
-		<edge from-layer="3311" from-port="0" to-layer="3312" to-port="2" />
-		<edge from-layer="3312" from-port="3" to-layer="3313" to-port="0" />
-		<edge from-layer="3313" from-port="2" to-layer="3314" to-port="0" />
-		<edge from-layer="3314" from-port="1" to-layer="3316" to-port="0" />
-		<edge from-layer="3315" from-port="0" to-layer="3316" to-port="1" />
-		<edge from-layer="3316" from-port="2" to-layer="3318" to-port="2" />
-		<edge from-layer="3317" from-port="0" to-layer="3318" to-port="3" />
-		<edge from-layer="3318" from-port="4" to-layer="3321" to-port="1" />
-		<edge from-layer="3318" from-port="4" to-layer="3327" to-port="2" />
-		<edge from-layer="3319" from-port="0" to-layer="3321" to-port="2" />
-		<edge from-layer="3320" from-port="0" to-layer="3321" to-port="3" />
-		<edge from-layer="3321" from-port="4" to-layer="3324" to-port="0" />
-		<edge from-layer="3322" from-port="0" to-layer="3323" to-port="0" />
-		<edge from-layer="3323" from-port="1" to-layer="3324" to-port="1" />
-		<edge from-layer="3324" from-port="2" to-layer="3328" to-port="0" />
-		<edge from-layer="3325" from-port="0" to-layer="3327" to-port="1" />
-		<edge from-layer="3326" from-port="0" to-layer="3327" to-port="3" />
-		<edge from-layer="3327" from-port="4" to-layer="3328" to-port="1" />
-		<edge from-layer="3328" from-port="2" to-layer="3337" to-port="0" />
-		<edge from-layer="3329" from-port="0" to-layer="3330" to-port="0" />
-		<edge from-layer="3330" from-port="1" to-layer="3333" to-port="0" />
-		<edge from-layer="3331" from-port="0" to-layer="3333" to-port="1" />
-		<edge from-layer="3332" from-port="0" to-layer="3333" to-port="3" />
-		<edge from-layer="3333" from-port="4" to-layer="3335" to-port="0" />
-		<edge from-layer="3334" from-port="0" to-layer="3335" to-port="2" />
-		<edge from-layer="3335" from-port="3" to-layer="3336" to-port="0" />
-		<edge from-layer="3336" from-port="2" to-layer="3337" to-port="1" />
-		<edge from-layer="3336" from-port="2" to-layer="3370" to-port="1" />
-		<edge from-layer="3337" from-port="2" to-layer="3338" to-port="1" />
-		<edge from-layer="3338" from-port="2" to-layer="3343" to-port="0" />
-		<edge from-layer="3339" from-port="0" to-layer="3342" to-port="1" />
-		<edge from-layer="3340" from-port="0" to-layer="3342" to-port="2" />
-		<edge from-layer="3341" from-port="0" to-layer="3342" to-port="3" />
-		<edge from-layer="3342" from-port="4" to-layer="3343" to-port="1" />
-		<edge from-layer="3343" from-port="2" to-layer="3401" to-port="0" />
-		<edge from-layer="3343" from-port="2" to-layer="3434" to-port="0" />
-		<edge from-layer="3344" from-port="0" to-layer="3347" to-port="1" />
-		<edge from-layer="3345" from-port="0" to-layer="3347" to-port="2" />
-		<edge from-layer="3346" from-port="0" to-layer="3347" to-port="3" />
-		<edge from-layer="3347" from-port="4" to-layer="3348" to-port="0" />
-		<edge from-layer="3347" from-port="4" to-layer="3351" to-port="0" />
-		<edge from-layer="3347" from-port="4" to-layer="3368" to-port="0" />
-		<edge from-layer="3347" from-port="4" to-layer="3362" to-port="0" />
-		<edge from-layer="3348" from-port="2" to-layer="3371" to-port="0" />
-		<edge from-layer="3349" from-port="0" to-layer="3359" to-port="0" />
-		<edge from-layer="3350" from-port="0" to-layer="3359" to-port="1" />
-		<edge from-layer="3351" from-port="1" to-layer="3353" to-port="0" />
-		<edge from-layer="3352" from-port="0" to-layer="3353" to-port="2" />
-		<edge from-layer="3353" from-port="3" to-layer="3354" to-port="0" />
-		<edge from-layer="3354" from-port="2" to-layer="3355" to-port="0" />
-		<edge from-layer="3355" from-port="1" to-layer="3357" to-port="0" />
-		<edge from-layer="3356" from-port="0" to-layer="3357" to-port="1" />
-		<edge from-layer="3357" from-port="2" to-layer="3359" to-port="2" />
-		<edge from-layer="3358" from-port="0" to-layer="3359" to-port="3" />
-		<edge from-layer="3359" from-port="4" to-layer="3368" to-port="2" />
-		<edge from-layer="3359" from-port="4" to-layer="3362" to-port="1" />
+		<edge from-layer="3267" from-port="4" to-layer="3268" to-port="1" />
+		<edge from-layer="3268" from-port="2" to-layer="3269" to-port="0" />
+		<edge from-layer="3269" from-port="1" to-layer="3271" to-port="0" />
+		<edge from-layer="3270" from-port="0" to-layer="3271" to-port="2" />
+		<edge from-layer="3271" from-port="3" to-layer="3278" to-port="0" />
+		<edge from-layer="3272" from-port="0" to-layer="3275" to-port="1" />
+		<edge from-layer="3273" from-port="0" to-layer="3275" to-port="2" />
+		<edge from-layer="3274" from-port="0" to-layer="3275" to-port="3" />
+		<edge from-layer="3275" from-port="5" to-layer="3277" to-port="0" />
+		<edge from-layer="3276" from-port="0" to-layer="3277" to-port="1" />
+		<edge from-layer="3277" from-port="2" to-layer="3278" to-port="1" />
+		<edge from-layer="3278" from-port="2" to-layer="3287" to-port="2" />
+		<edge from-layer="3278" from-port="2" to-layer="5136" to-port="0" />
+		<edge from-layer="3279" from-port="0" to-layer="3286" to-port="1" />
+		<edge from-layer="3280" from-port="1" to-layer="3283" to-port="0" />
+		<edge from-layer="3281" from-port="0" to-layer="3283" to-port="1" />
+		<edge from-layer="3282" from-port="0" to-layer="3283" to-port="2" />
+		<edge from-layer="3283" from-port="3" to-layer="3284" to-port="0" />
+		<edge from-layer="3284" from-port="2" to-layer="3286" to-port="2" />
+		<edge from-layer="3285" from-port="0" to-layer="3286" to-port="3" />
+		<edge from-layer="3286" from-port="5" to-layer="3287" to-port="3" />
+		<edge from-layer="3287" from-port="4" to-layer="3289" to-port="0" />
+		<edge from-layer="3288" from-port="0" to-layer="3289" to-port="1" />
+		<edge from-layer="3289" from-port="2" to-layer="3291" to-port="0" />
+		<edge from-layer="3290" from-port="0" to-layer="3291" to-port="1" />
+		<edge from-layer="3291" from-port="2" to-layer="3300" to-port="0" />
+		<edge from-layer="3292" from-port="0" to-layer="3293" to-port="0" />
+		<edge from-layer="3293" from-port="1" to-layer="3296" to-port="0" />
+		<edge from-layer="3294" from-port="0" to-layer="3295" to-port="0" />
+		<edge from-layer="3295" from-port="1" to-layer="3296" to-port="1" />
+		<edge from-layer="3296" from-port="2" to-layer="3298" to-port="0" />
+		<edge from-layer="3297" from-port="0" to-layer="3298" to-port="1" />
+		<edge from-layer="3298" from-port="2" to-layer="3299" to-port="0" />
+		<edge from-layer="3299" from-port="1" to-layer="3300" to-port="1" />
+		<edge from-layer="3300" from-port="2" to-layer="3302" to-port="0" />
+		<edge from-layer="3301" from-port="0" to-layer="3302" to-port="1" />
+		<edge from-layer="3302" from-port="2" to-layer="3303" to-port="1" />
+		<edge from-layer="3303" from-port="2" to-layer="3304" to-port="0" />
+		<edge from-layer="3304" from-port="2" to-layer="3462" to-port="1" />
+		<edge from-layer="3304" from-port="2" to-layer="3306" to-port="0" />
+		<edge from-layer="3305" from-port="0" to-layer="3306" to-port="1" />
+		<edge from-layer="3306" from-port="2" to-layer="3337" to-port="0" />
+		<edge from-layer="3306" from-port="2" to-layer="3308" to-port="0" />
+		<edge from-layer="3307" from-port="0" to-layer="3308" to-port="1" />
+		<edge from-layer="3308" from-port="2" to-layer="3310" to-port="0" />
+		<edge from-layer="3309" from-port="0" to-layer="3310" to-port="1" />
+		<edge from-layer="3310" from-port="2" to-layer="3319" to-port="0" />
+		<edge from-layer="3311" from-port="0" to-layer="3312" to-port="0" />
+		<edge from-layer="3312" from-port="1" to-layer="3315" to-port="0" />
+		<edge from-layer="3313" from-port="0" to-layer="3314" to-port="0" />
+		<edge from-layer="3314" from-port="1" to-layer="3315" to-port="1" />
+		<edge from-layer="3315" from-port="2" to-layer="3317" to-port="0" />
+		<edge from-layer="3316" from-port="0" to-layer="3317" to-port="1" />
+		<edge from-layer="3317" from-port="2" to-layer="3318" to-port="0" />
+		<edge from-layer="3318" from-port="1" to-layer="3319" to-port="1" />
+		<edge from-layer="3319" from-port="2" to-layer="3322" to-port="0" />
+		<edge from-layer="3320" from-port="0" to-layer="3321" to-port="0" />
+		<edge from-layer="3321" from-port="1" to-layer="3322" to-port="1" />
+		<edge from-layer="3322" from-port="2" to-layer="3323" to-port="0" />
+		<edge from-layer="3323" from-port="1" to-layer="3332" to-port="0" />
+		<edge from-layer="3324" from-port="0" to-layer="3325" to-port="0" />
+		<edge from-layer="3325" from-port="1" to-layer="3328" to-port="0" />
+		<edge from-layer="3326" from-port="0" to-layer="3327" to-port="0" />
+		<edge from-layer="3327" from-port="1" to-layer="3328" to-port="1" />
+		<edge from-layer="3328" from-port="2" to-layer="3330" to-port="0" />
+		<edge from-layer="3329" from-port="0" to-layer="3330" to-port="1" />
+		<edge from-layer="3330" from-port="2" to-layer="3331" to-port="0" />
+		<edge from-layer="3331" from-port="1" to-layer="3332" to-port="1" />
+		<edge from-layer="3332" from-port="2" to-layer="3335" to-port="0" />
+		<edge from-layer="3333" from-port="0" to-layer="3334" to-port="0" />
+		<edge from-layer="3334" from-port="1" to-layer="3335" to-port="1" />
+		<edge from-layer="3335" from-port="2" to-layer="3461" to-port="0" />
+		<edge from-layer="3336" from-port="0" to-layer="3337" to-port="1" />
+		<edge from-layer="3337" from-port="2" to-layer="3339" to-port="0" />
+		<edge from-layer="3338" from-port="0" to-layer="3339" to-port="1" />
+		<edge from-layer="3339" from-port="2" to-layer="3348" to-port="0" />
+		<edge from-layer="3340" from-port="0" to-layer="3341" to-port="0" />
+		<edge from-layer="3341" from-port="1" to-layer="3344" to-port="0" />
+		<edge from-layer="3342" from-port="0" to-layer="3343" to-port="0" />
+		<edge from-layer="3343" from-port="1" to-layer="3344" to-port="1" />
+		<edge from-layer="3344" from-port="2" to-layer="3346" to-port="0" />
+		<edge from-layer="3345" from-port="0" to-layer="3346" to-port="1" />
+		<edge from-layer="3346" from-port="2" to-layer="3347" to-port="0" />
+		<edge from-layer="3347" from-port="1" to-layer="3348" to-port="1" />
+		<edge from-layer="3348" from-port="2" to-layer="3350" to-port="0" />
+		<edge from-layer="3349" from-port="0" to-layer="3350" to-port="1" />
+		<edge from-layer="3350" from-port="2" to-layer="3352" to-port="0" />
+		<edge from-layer="3351" from-port="0" to-layer="3352" to-port="1" />
+		<edge from-layer="3352" from-port="2" to-layer="3433" to-port="0" />
+		<edge from-layer="3352" from-port="2" to-layer="3356" to-port="0" />
+		<edge from-layer="3352" from-port="2" to-layer="3394" to-port="0" />
+		<edge from-layer="3353" from-port="0" to-layer="3356" to-port="1" />
+		<edge from-layer="3354" from-port="0" to-layer="3356" to-port="2" />
+		<edge from-layer="3355" from-port="0" to-layer="3356" to-port="3" />
+		<edge from-layer="3356" from-port="5" to-layer="3358" to-port="0" />
+		<edge from-layer="3357" from-port="0" to-layer="3358" to-port="1" />
+		<edge from-layer="3358" from-port="2" to-layer="3362" to-port="0" />
+		<edge from-layer="3358" from-port="2" to-layer="3380" to-port="0" />
+		<edge from-layer="3359" from-port="0" to-layer="3362" to-port="1" />
 		<edge from-layer="3360" from-port="0" to-layer="3362" to-port="2" />
 		<edge from-layer="3361" from-port="0" to-layer="3362" to-port="3" />
-		<edge from-layer="3362" from-port="4" to-layer="3365" to-port="0" />
-		<edge from-layer="3363" from-port="0" to-layer="3364" to-port="0" />
-		<edge from-layer="3364" from-port="1" to-layer="3365" to-port="1" />
-		<edge from-layer="3365" from-port="2" to-layer="3369" to-port="0" />
-		<edge from-layer="3366" from-port="0" to-layer="3368" to-port="1" />
-		<edge from-layer="3367" from-port="0" to-layer="3368" to-port="3" />
-		<edge from-layer="3368" from-port="4" to-layer="3369" to-port="1" />
-		<edge from-layer="3369" from-port="2" to-layer="3370" to-port="0" />
-		<edge from-layer="3370" from-port="2" to-layer="3371" to-port="1" />
-		<edge from-layer="3371" from-port="2" to-layer="3376" to-port="0" />
-		<edge from-layer="3372" from-port="0" to-layer="3375" to-port="1" />
-		<edge from-layer="3373" from-port="0" to-layer="3375" to-port="2" />
-		<edge from-layer="3374" from-port="0" to-layer="3375" to-port="3" />
-		<edge from-layer="3375" from-port="4" to-layer="3376" to-port="1" />
-		<edge from-layer="3376" from-port="2" to-layer="3377" to-port="1" />
-		<edge from-layer="3377" from-port="2" to-layer="3434" to-port="1" />
-		<edge from-layer="3377" from-port="2" to-layer="3398" to-port="0" />
-		<edge from-layer="3377" from-port="2" to-layer="7712" to-port="0" />
-		<edge from-layer="3378" from-port="0" to-layer="3379" to-port="0" />
-		<edge from-layer="3379" from-port="1" to-layer="3384" to-port="0" />
-		<edge from-layer="3380" from-port="0" to-layer="3383" to-port="1" />
-		<edge from-layer="3381" from-port="0" to-layer="3383" to-port="2" />
-		<edge from-layer="3382" from-port="0" to-layer="3383" to-port="3" />
-		<edge from-layer="3383" from-port="4" to-layer="3384" to-port="1" />
-		<edge from-layer="3384" from-port="2" to-layer="3385" to-port="0" />
-		<edge from-layer="3385" from-port="1" to-layer="3387" to-port="0" />
-		<edge from-layer="3386" from-port="0" to-layer="3387" to-port="2" />
-		<edge from-layer="3387" from-port="3" to-layer="3394" to-port="0" />
-		<edge from-layer="3388" from-port="0" to-layer="3391" to-port="1" />
-		<edge from-layer="3389" from-port="0" to-layer="3391" to-port="2" />
-		<edge from-layer="3390" from-port="0" to-layer="3391" to-port="3" />
-		<edge from-layer="3391" from-port="4" to-layer="3393" to-port="0" />
-		<edge from-layer="3392" from-port="0" to-layer="3393" to-port="1" />
-		<edge from-layer="3393" from-port="2" to-layer="3394" to-port="1" />
-		<edge from-layer="3394" from-port="2" to-layer="3434" to-port="2" />
-		<edge from-layer="3394" from-port="2" to-layer="7711" to-port="0" />
-		<edge from-layer="3395" from-port="0" to-layer="3413" to-port="0" />
-		<edge from-layer="3396" from-port="0" to-layer="3408" to-port="0" />
-		<edge from-layer="3397" from-port="0" to-layer="3411" to-port="1" />
-		<edge from-layer="3397" from-port="0" to-layer="3408" to-port="1" />
-		<edge from-layer="3398" from-port="1" to-layer="3400" to-port="0" />
-		<edge from-layer="3399" from-port="0" to-layer="3400" to-port="2" />
-		<edge from-layer="3400" from-port="3" to-layer="3404" to-port="0" />
-		<edge from-layer="3400" from-port="3" to-layer="3410" to-port="0" />
-		<edge from-layer="3401" from-port="1" to-layer="3403" to-port="0" />
-		<edge from-layer="3401" from-port="1" to-layer="3429" to-port="0" />
-		<edge from-layer="3402" from-port="0" to-layer="3403" to-port="2" />
-		<edge from-layer="3403" from-port="3" to-layer="3404" to-port="1" />
-		<edge from-layer="3404" from-port="2" to-layer="3406" to-port="0" />
-		<edge from-layer="3405" from-port="0" to-layer="3406" to-port="1" />
-		<edge from-layer="3405" from-port="0" to-layer="3410" to-port="1" />
-		<edge from-layer="3406" from-port="2" to-layer="3408" to-port="2" />
-		<edge from-layer="3407" from-port="0" to-layer="3408" to-port="3" />
-		<edge from-layer="3407" from-port="0" to-layer="3411" to-port="3" />
-		<edge from-layer="3408" from-port="4" to-layer="3413" to-port="1" />
-		<edge from-layer="3409" from-port="0" to-layer="3411" to-port="0" />
-		<edge from-layer="3410" from-port="2" to-layer="3411" to-port="2" />
-		<edge from-layer="3410" from-port="2" to-layer="3418" to-port="2" />
-		<edge from-layer="3411" from-port="4" to-layer="3413" to-port="2" />
-		<edge from-layer="3412" from-port="0" to-layer="3413" to-port="3" />
-		<edge from-layer="3413" from-port="4" to-layer="3420" to-port="0" />
-		<edge from-layer="3414" from-port="0" to-layer="3420" to-port="1" />
-		<edge from-layer="3415" from-port="0" to-layer="3418" to-port="0" />
-		<edge from-layer="3416" from-port="0" to-layer="3418" to-port="1" />
+		<edge from-layer="3362" from-port="5" to-layer="3367" to-port="0" />
+		<edge from-layer="3362" from-port="5" to-layer="3373" to-port="0" />
+		<edge from-layer="3362" from-port="5" to-layer="3363" to-port="0" />
+		<edge from-layer="3363" from-port="2" to-layer="3376" to-port="0" />
+		<edge from-layer="3364" from-port="0" to-layer="3367" to-port="1" />
+		<edge from-layer="3365" from-port="0" to-layer="3367" to-port="2" />
+		<edge from-layer="3366" from-port="0" to-layer="3367" to-port="3" />
+		<edge from-layer="3367" from-port="5" to-layer="3369" to-port="0" />
+		<edge from-layer="3368" from-port="0" to-layer="3369" to-port="1" />
+		<edge from-layer="3369" from-port="2" to-layer="3374" to-port="0" />
+		<edge from-layer="3370" from-port="0" to-layer="3373" to-port="1" />
+		<edge from-layer="3371" from-port="0" to-layer="3373" to-port="2" />
+		<edge from-layer="3372" from-port="0" to-layer="3373" to-port="3" />
+		<edge from-layer="3373" from-port="5" to-layer="3374" to-port="1" />
+		<edge from-layer="3374" from-port="2" to-layer="3375" to-port="0" />
+		<edge from-layer="3375" from-port="2" to-layer="3376" to-port="1" />
+		<edge from-layer="3376" from-port="2" to-layer="3381" to-port="0" />
+		<edge from-layer="3377" from-port="0" to-layer="3380" to-port="1" />
+		<edge from-layer="3378" from-port="0" to-layer="3380" to-port="2" />
+		<edge from-layer="3379" from-port="0" to-layer="3380" to-port="3" />
+		<edge from-layer="3380" from-port="5" to-layer="3381" to-port="1" />
+		<edge from-layer="3381" from-port="2" to-layer="3445" to-port="0" />
+		<edge from-layer="3382" from-port="0" to-layer="3387" to-port="0" />
+		<edge from-layer="3383" from-port="0" to-layer="3386" to-port="1" />
+		<edge from-layer="3384" from-port="0" to-layer="3386" to-port="2" />
+		<edge from-layer="3385" from-port="0" to-layer="3386" to-port="3" />
+		<edge from-layer="3386" from-port="4" to-layer="3387" to-port="1" />
+		<edge from-layer="3387" from-port="2" to-layer="3388" to-port="0" />
+		<edge from-layer="3388" from-port="1" to-layer="3390" to-port="0" />
+		<edge from-layer="3389" from-port="0" to-layer="3390" to-port="2" />
+		<edge from-layer="3390" from-port="3" to-layer="3420" to-port="0" />
+		<edge from-layer="3390" from-port="3" to-layer="3438" to-port="0" />
+		<edge from-layer="3391" from-port="0" to-layer="3394" to-port="1" />
+		<edge from-layer="3392" from-port="0" to-layer="3394" to-port="2" />
+		<edge from-layer="3393" from-port="0" to-layer="3394" to-port="3" />
+		<edge from-layer="3394" from-port="5" to-layer="3396" to-port="0" />
+		<edge from-layer="3395" from-port="0" to-layer="3396" to-port="1" />
+		<edge from-layer="3396" from-port="2" to-layer="3400" to-port="0" />
+		<edge from-layer="3396" from-port="2" to-layer="3418" to-port="0" />
+		<edge from-layer="3397" from-port="0" to-layer="3400" to-port="1" />
+		<edge from-layer="3398" from-port="0" to-layer="3400" to-port="2" />
+		<edge from-layer="3399" from-port="0" to-layer="3400" to-port="3" />
+		<edge from-layer="3400" from-port="5" to-layer="3401" to-port="0" />
+		<edge from-layer="3400" from-port="5" to-layer="3405" to-port="0" />
+		<edge from-layer="3400" from-port="5" to-layer="3411" to-port="0" />
+		<edge from-layer="3401" from-port="2" to-layer="3414" to-port="0" />
+		<edge from-layer="3402" from-port="0" to-layer="3405" to-port="1" />
+		<edge from-layer="3403" from-port="0" to-layer="3405" to-port="2" />
+		<edge from-layer="3404" from-port="0" to-layer="3405" to-port="3" />
+		<edge from-layer="3405" from-port="5" to-layer="3407" to-port="0" />
+		<edge from-layer="3406" from-port="0" to-layer="3407" to-port="1" />
+		<edge from-layer="3407" from-port="2" to-layer="3412" to-port="0" />
+		<edge from-layer="3408" from-port="0" to-layer="3411" to-port="1" />
+		<edge from-layer="3409" from-port="0" to-layer="3411" to-port="2" />
+		<edge from-layer="3410" from-port="0" to-layer="3411" to-port="3" />
+		<edge from-layer="3411" from-port="5" to-layer="3412" to-port="1" />
+		<edge from-layer="3412" from-port="2" to-layer="3413" to-port="0" />
+		<edge from-layer="3413" from-port="2" to-layer="3414" to-port="1" />
+		<edge from-layer="3414" from-port="2" to-layer="3419" to-port="0" />
+		<edge from-layer="3415" from-port="0" to-layer="3418" to-port="1" />
+		<edge from-layer="3416" from-port="0" to-layer="3418" to-port="2" />
 		<edge from-layer="3417" from-port="0" to-layer="3418" to-port="3" />
-		<edge from-layer="3418" from-port="4" to-layer="3420" to-port="2" />
-		<edge from-layer="3419" from-port="0" to-layer="3420" to-port="3" />
-		<edge from-layer="3420" from-port="4" to-layer="3424" to-port="0" />
-		<edge from-layer="3421" from-port="0" to-layer="3422" to-port="0" />
-		<edge from-layer="3422" from-port="1" to-layer="3424" to-port="1" />
-		<edge from-layer="3423" from-port="0" to-layer="3424" to-port="2" />
-		<edge from-layer="3424" from-port="3" to-layer="3425" to-port="0" />
-		<edge from-layer="3425" from-port="2" to-layer="3426" to-port="0" />
-		<edge from-layer="3425" from-port="2" to-layer="3433" to-port="0" />
-		<edge from-layer="3426" from-port="1" to-layer="3432" to-port="0" />
-		<edge from-layer="3427" from-port="0" to-layer="3429" to-port="1" />
+		<edge from-layer="3418" from-port="5" to-layer="3419" to-port="1" />
+		<edge from-layer="3419" from-port="2" to-layer="3420" to-port="1" />
+		<edge from-layer="3420" from-port="2" to-layer="5133" to-port="0" />
+		<edge from-layer="3420" from-port="2" to-layer="3445" to-port="1" />
+		<edge from-layer="3421" from-port="0" to-layer="3426" to-port="0" />
+		<edge from-layer="3422" from-port="0" to-layer="3425" to-port="1" />
+		<edge from-layer="3423" from-port="0" to-layer="3425" to-port="2" />
+		<edge from-layer="3424" from-port="0" to-layer="3425" to-port="3" />
+		<edge from-layer="3425" from-port="4" to-layer="3426" to-port="1" />
+		<edge from-layer="3426" from-port="2" to-layer="3427" to-port="0" />
+		<edge from-layer="3427" from-port="1" to-layer="3429" to-port="0" />
 		<edge from-layer="3428" from-port="0" to-layer="3429" to-port="2" />
-		<edge from-layer="3429" from-port="3" to-layer="3430" to-port="0" />
-		<edge from-layer="3430" from-port="4" to-layer="3431" to-port="0" />
-		<edge from-layer="3431" from-port="1" to-layer="3432" to-port="1" />
-		<edge from-layer="3432" from-port="2" to-layer="3433" to-port="1" />
-		<edge from-layer="3433" from-port="2" to-layer="3434" to-port="3" />
-		<edge from-layer="3434" from-port="4" to-layer="3436" to-port="0" />
-		<edge from-layer="3435" from-port="0" to-layer="3436" to-port="1" />
-		<edge from-layer="3436" from-port="2" to-layer="3438" to-port="0" />
-		<edge from-layer="3437" from-port="0" to-layer="3438" to-port="1" />
-		<edge from-layer="3438" from-port="2" to-layer="3447" to-port="0" />
-		<edge from-layer="3439" from-port="0" to-layer="3440" to-port="0" />
-		<edge from-layer="3440" from-port="1" to-layer="3443" to-port="0" />
-		<edge from-layer="3441" from-port="0" to-layer="3442" to-port="0" />
-		<edge from-layer="3442" from-port="1" to-layer="3443" to-port="1" />
-		<edge from-layer="3443" from-port="2" to-layer="3445" to-port="0" />
-		<edge from-layer="3444" from-port="0" to-layer="3445" to-port="1" />
-		<edge from-layer="3445" from-port="2" to-layer="3446" to-port="0" />
-		<edge from-layer="3446" from-port="1" to-layer="3447" to-port="1" />
-		<edge from-layer="3447" from-port="2" to-layer="3450" to-port="0" />
-		<edge from-layer="3448" from-port="0" to-layer="3449" to-port="0" />
-		<edge from-layer="3449" from-port="1" to-layer="3450" to-port="1" />
-		<edge from-layer="3450" from-port="2" to-layer="3451" to-port="1" />
-		<edge from-layer="3451" from-port="2" to-layer="3452" to-port="0" />
-		<edge from-layer="3452" from-port="2" to-layer="3454" to-port="0" />
-		<edge from-layer="3452" from-port="2" to-layer="3694" to-port="1" />
-		<edge from-layer="3453" from-port="0" to-layer="3454" to-port="1" />
-		<edge from-layer="3454" from-port="2" to-layer="3488" to-port="0" />
-		<edge from-layer="3454" from-port="2" to-layer="3457" to-port="0" />
-		<edge from-layer="3455" from-port="0" to-layer="3456" to-port="0" />
-		<edge from-layer="3456" from-port="1" to-layer="3457" to-port="1" />
-		<edge from-layer="3457" from-port="2" to-layer="3460" to-port="0" />
-		<edge from-layer="3458" from-port="0" to-layer="3459" to-port="0" />
-		<edge from-layer="3459" from-port="1" to-layer="3460" to-port="1" />
-		<edge from-layer="3460" from-port="2" to-layer="3469" to-port="0" />
-		<edge from-layer="3461" from-port="0" to-layer="3462" to-port="0" />
-		<edge from-layer="3462" from-port="1" to-layer="3465" to-port="0" />
-		<edge from-layer="3463" from-port="0" to-layer="3464" to-port="0" />
-		<edge from-layer="3464" from-port="1" to-layer="3465" to-port="1" />
-		<edge from-layer="3465" from-port="2" to-layer="3467" to-port="0" />
-		<edge from-layer="3466" from-port="0" to-layer="3467" to-port="1" />
-		<edge from-layer="3467" from-port="2" to-layer="3468" to-port="0" />
-		<edge from-layer="3468" from-port="1" to-layer="3469" to-port="1" />
-		<edge from-layer="3469" from-port="2" to-layer="3472" to-port="0" />
-		<edge from-layer="3470" from-port="0" to-layer="3471" to-port="0" />
-		<edge from-layer="3471" from-port="1" to-layer="3472" to-port="1" />
-		<edge from-layer="3472" from-port="2" to-layer="3473" to-port="0" />
-		<edge from-layer="3473" from-port="1" to-layer="3482" to-port="0" />
-		<edge from-layer="3474" from-port="0" to-layer="3475" to-port="0" />
-		<edge from-layer="3475" from-port="1" to-layer="3478" to-port="0" />
-		<edge from-layer="3476" from-port="0" to-layer="3477" to-port="0" />
-		<edge from-layer="3477" from-port="1" to-layer="3478" to-port="1" />
-		<edge from-layer="3478" from-port="2" to-layer="3480" to-port="0" />
-		<edge from-layer="3479" from-port="0" to-layer="3480" to-port="1" />
+		<edge from-layer="3429" from-port="3" to-layer="3436" to-port="0" />
+		<edge from-layer="3430" from-port="0" to-layer="3433" to-port="1" />
+		<edge from-layer="3431" from-port="0" to-layer="3433" to-port="2" />
+		<edge from-layer="3432" from-port="0" to-layer="3433" to-port="3" />
+		<edge from-layer="3433" from-port="5" to-layer="3435" to-port="0" />
+		<edge from-layer="3434" from-port="0" to-layer="3435" to-port="1" />
+		<edge from-layer="3435" from-port="2" to-layer="3436" to-port="1" />
+		<edge from-layer="3436" from-port="2" to-layer="5132" to-port="0" />
+		<edge from-layer="3436" from-port="2" to-layer="3445" to-port="2" />
+		<edge from-layer="3437" from-port="0" to-layer="3444" to-port="1" />
+		<edge from-layer="3438" from-port="1" to-layer="3441" to-port="0" />
+		<edge from-layer="3439" from-port="0" to-layer="3441" to-port="1" />
+		<edge from-layer="3440" from-port="0" to-layer="3441" to-port="2" />
+		<edge from-layer="3441" from-port="3" to-layer="3442" to-port="0" />
+		<edge from-layer="3442" from-port="2" to-layer="3444" to-port="2" />
+		<edge from-layer="3443" from-port="0" to-layer="3444" to-port="3" />
+		<edge from-layer="3444" from-port="5" to-layer="3445" to-port="3" />
+		<edge from-layer="3445" from-port="4" to-layer="3447" to-port="0" />
+		<edge from-layer="3446" from-port="0" to-layer="3447" to-port="1" />
+		<edge from-layer="3447" from-port="2" to-layer="3449" to-port="0" />
+		<edge from-layer="3448" from-port="0" to-layer="3449" to-port="1" />
+		<edge from-layer="3449" from-port="2" to-layer="3458" to-port="0" />
+		<edge from-layer="3450" from-port="0" to-layer="3451" to-port="0" />
+		<edge from-layer="3451" from-port="1" to-layer="3454" to-port="0" />
+		<edge from-layer="3452" from-port="0" to-layer="3453" to-port="0" />
+		<edge from-layer="3453" from-port="1" to-layer="3454" to-port="1" />
+		<edge from-layer="3454" from-port="2" to-layer="3456" to-port="0" />
+		<edge from-layer="3455" from-port="0" to-layer="3456" to-port="1" />
+		<edge from-layer="3456" from-port="2" to-layer="3457" to-port="0" />
+		<edge from-layer="3457" from-port="1" to-layer="3458" to-port="1" />
+		<edge from-layer="3458" from-port="2" to-layer="3460" to-port="0" />
+		<edge from-layer="3459" from-port="0" to-layer="3460" to-port="1" />
+		<edge from-layer="3460" from-port="2" to-layer="3461" to-port="1" />
+		<edge from-layer="3461" from-port="2" to-layer="3462" to-port="0" />
+		<edge from-layer="3462" from-port="2" to-layer="3620" to-port="1" />
+		<edge from-layer="3462" from-port="2" to-layer="3464" to-port="0" />
+		<edge from-layer="3463" from-port="0" to-layer="3464" to-port="1" />
+		<edge from-layer="3464" from-port="2" to-layer="3495" to-port="0" />
+		<edge from-layer="3464" from-port="2" to-layer="3466" to-port="0" />
+		<edge from-layer="3465" from-port="0" to-layer="3466" to-port="1" />
+		<edge from-layer="3466" from-port="2" to-layer="3468" to-port="0" />
+		<edge from-layer="3467" from-port="0" to-layer="3468" to-port="1" />
+		<edge from-layer="3468" from-port="2" to-layer="3477" to-port="0" />
+		<edge from-layer="3469" from-port="0" to-layer="3470" to-port="0" />
+		<edge from-layer="3470" from-port="1" to-layer="3473" to-port="0" />
+		<edge from-layer="3471" from-port="0" to-layer="3472" to-port="0" />
+		<edge from-layer="3472" from-port="1" to-layer="3473" to-port="1" />
+		<edge from-layer="3473" from-port="2" to-layer="3475" to-port="0" />
+		<edge from-layer="3474" from-port="0" to-layer="3475" to-port="1" />
+		<edge from-layer="3475" from-port="2" to-layer="3476" to-port="0" />
+		<edge from-layer="3476" from-port="1" to-layer="3477" to-port="1" />
+		<edge from-layer="3477" from-port="2" to-layer="3480" to-port="0" />
+		<edge from-layer="3478" from-port="0" to-layer="3479" to-port="0" />
+		<edge from-layer="3479" from-port="1" to-layer="3480" to-port="1" />
 		<edge from-layer="3480" from-port="2" to-layer="3481" to-port="0" />
-		<edge from-layer="3481" from-port="1" to-layer="3482" to-port="1" />
-		<edge from-layer="3482" from-port="2" to-layer="3485" to-port="0" />
-		<edge from-layer="3483" from-port="0" to-layer="3484" to-port="0" />
-		<edge from-layer="3484" from-port="1" to-layer="3485" to-port="1" />
-		<edge from-layer="3485" from-port="2" to-layer="3693" to-port="0" />
-		<edge from-layer="3486" from-port="0" to-layer="3487" to-port="0" />
-		<edge from-layer="3487" from-port="1" to-layer="3488" to-port="1" />
-		<edge from-layer="3488" from-port="2" to-layer="3491" to-port="0" />
-		<edge from-layer="3489" from-port="0" to-layer="3490" to-port="0" />
-		<edge from-layer="3490" from-port="1" to-layer="3491" to-port="1" />
-		<edge from-layer="3491" from-port="2" to-layer="3500" to-port="0" />
-		<edge from-layer="3492" from-port="0" to-layer="3493" to-port="0" />
-		<edge from-layer="3493" from-port="1" to-layer="3496" to-port="0" />
-		<edge from-layer="3494" from-port="0" to-layer="3495" to-port="0" />
-		<edge from-layer="3495" from-port="1" to-layer="3496" to-port="1" />
-		<edge from-layer="3496" from-port="2" to-layer="3498" to-port="0" />
-		<edge from-layer="3497" from-port="0" to-layer="3498" to-port="1" />
-		<edge from-layer="3498" from-port="2" to-layer="3499" to-port="0" />
-		<edge from-layer="3499" from-port="1" to-layer="3500" to-port="1" />
-		<edge from-layer="3500" from-port="2" to-layer="3503" to-port="0" />
-		<edge from-layer="3501" from-port="0" to-layer="3502" to-port="0" />
-		<edge from-layer="3502" from-port="1" to-layer="3503" to-port="1" />
-		<edge from-layer="3503" from-port="2" to-layer="3505" to-port="0" />
-		<edge from-layer="3504" from-port="0" to-layer="3505" to-port="1" />
-		<edge from-layer="3505" from-port="2" to-layer="3509" to-port="0" />
-		<edge from-layer="3505" from-port="2" to-layer="3522" to-port="0" />
-		<edge from-layer="3505" from-port="2" to-layer="3633" to-port="0" />
-		<edge from-layer="3506" from-port="0" to-layer="3509" to-port="1" />
-		<edge from-layer="3507" from-port="0" to-layer="3509" to-port="2" />
-		<edge from-layer="3508" from-port="0" to-layer="3509" to-port="3" />
-		<edge from-layer="3509" from-port="4" to-layer="3511" to-port="0" />
-		<edge from-layer="3510" from-port="0" to-layer="3511" to-port="1" />
-		<edge from-layer="3511" from-port="2" to-layer="3515" to-port="0" />
-		<edge from-layer="3511" from-port="2" to-layer="3584" to-port="0" />
-		<edge from-layer="3512" from-port="0" to-layer="3515" to-port="1" />
-		<edge from-layer="3513" from-port="0" to-layer="3515" to-port="2" />
-		<edge from-layer="3514" from-port="0" to-layer="3515" to-port="3" />
-		<edge from-layer="3515" from-port="4" to-layer="3552" to-port="0" />
-		<edge from-layer="3515" from-port="4" to-layer="3569" to-port="0" />
-		<edge from-layer="3515" from-port="4" to-layer="3563" to-port="0" />
-		<edge from-layer="3515" from-port="4" to-layer="3549" to-port="0" />
-		<edge from-layer="3516" from-port="0" to-layer="3517" to-port="0" />
-		<edge from-layer="3517" from-port="1" to-layer="3545" to-port="0" />
-		<edge from-layer="3518" from-port="0" to-layer="3545" to-port="1" />
-		<edge from-layer="3519" from-port="0" to-layer="3522" to-port="1" />
-		<edge from-layer="3520" from-port="0" to-layer="3522" to-port="2" />
-		<edge from-layer="3521" from-port="0" to-layer="3522" to-port="3" />
-		<edge from-layer="3522" from-port="4" to-layer="3524" to-port="0" />
-		<edge from-layer="3523" from-port="0" to-layer="3524" to-port="1" />
-		<edge from-layer="3524" from-port="2" to-layer="3525" to-port="0" />
-		<edge from-layer="3524" from-port="2" to-layer="3589" to-port="0" />
-		<edge from-layer="3524" from-port="2" to-layer="3617" to-port="0" />
-		<edge from-layer="3525" from-port="1" to-layer="3527" to-port="0" />
-		<edge from-layer="3526" from-port="0" to-layer="3527" to-port="2" />
-		<edge from-layer="3527" from-port="3" to-layer="3541" to-port="0" />
-		<edge from-layer="3528" from-port="0" to-layer="3529" to-port="0" />
-		<edge from-layer="3529" from-port="1" to-layer="3534" to-port="0" />
-		<edge from-layer="3530" from-port="0" to-layer="3533" to-port="1" />
-		<edge from-layer="3531" from-port="0" to-layer="3533" to-port="2" />
-		<edge from-layer="3532" from-port="0" to-layer="3533" to-port="3" />
-		<edge from-layer="3533" from-port="4" to-layer="3534" to-port="1" />
-		<edge from-layer="3534" from-port="2" to-layer="3535" to-port="0" />
-		<edge from-layer="3535" from-port="1" to-layer="3537" to-port="0" />
-		<edge from-layer="3536" from-port="0" to-layer="3537" to-port="2" />
-		<edge from-layer="3537" from-port="3" to-layer="3619" to-port="0" />
-		<edge from-layer="3537" from-port="3" to-layer="3538" to-port="0" />
-		<edge from-layer="3538" from-port="1" to-layer="3540" to-port="0" />
-		<edge from-layer="3539" from-port="0" to-layer="3540" to-port="2" />
-		<edge from-layer="3540" from-port="3" to-layer="3541" to-port="1" />
-		<edge from-layer="3541" from-port="2" to-layer="3543" to-port="0" />
-		<edge from-layer="3542" from-port="0" to-layer="3543" to-port="1" />
-		<edge from-layer="3543" from-port="2" to-layer="3545" to-port="2" />
-		<edge from-layer="3543" from-port="2" to-layer="3575" to-port="2" />
-		<edge from-layer="3544" from-port="0" to-layer="3545" to-port="3" />
-		<edge from-layer="3545" from-port="4" to-layer="3547" to-port="0" />
-		<edge from-layer="3546" from-port="0" to-layer="3547" to-port="2" />
-		<edge from-layer="3547" from-port="3" to-layer="3548" to-port="0" />
-		<edge from-layer="3548" from-port="2" to-layer="3590" to-port="1" />
-		<edge from-layer="3548" from-port="2" to-layer="3549" to-port="1" />
-		<edge from-layer="3549" from-port="2" to-layer="3580" to-port="0" />
-		<edge from-layer="3550" from-port="0" to-layer="3560" to-port="0" />
-		<edge from-layer="3551" from-port="0" to-layer="3560" to-port="1" />
-		<edge from-layer="3552" from-port="1" to-layer="3554" to-port="0" />
-		<edge from-layer="3553" from-port="0" to-layer="3554" to-port="2" />
-		<edge from-layer="3554" from-port="3" to-layer="3555" to-port="0" />
-		<edge from-layer="3555" from-port="2" to-layer="3556" to-port="0" />
-		<edge from-layer="3556" from-port="1" to-layer="3558" to-port="0" />
-		<edge from-layer="3557" from-port="0" to-layer="3558" to-port="1" />
-		<edge from-layer="3558" from-port="2" to-layer="3560" to-port="2" />
-		<edge from-layer="3559" from-port="0" to-layer="3560" to-port="3" />
-		<edge from-layer="3560" from-port="4" to-layer="3569" to-port="2" />
-		<edge from-layer="3560" from-port="4" to-layer="3563" to-port="1" />
+		<edge from-layer="3481" from-port="1" to-layer="3490" to-port="0" />
+		<edge from-layer="3482" from-port="0" to-layer="3483" to-port="0" />
+		<edge from-layer="3483" from-port="1" to-layer="3486" to-port="0" />
+		<edge from-layer="3484" from-port="0" to-layer="3485" to-port="0" />
+		<edge from-layer="3485" from-port="1" to-layer="3486" to-port="1" />
+		<edge from-layer="3486" from-port="2" to-layer="3488" to-port="0" />
+		<edge from-layer="3487" from-port="0" to-layer="3488" to-port="1" />
+		<edge from-layer="3488" from-port="2" to-layer="3489" to-port="0" />
+		<edge from-layer="3489" from-port="1" to-layer="3490" to-port="1" />
+		<edge from-layer="3490" from-port="2" to-layer="3493" to-port="0" />
+		<edge from-layer="3491" from-port="0" to-layer="3492" to-port="0" />
+		<edge from-layer="3492" from-port="1" to-layer="3493" to-port="1" />
+		<edge from-layer="3493" from-port="2" to-layer="3619" to-port="0" />
+		<edge from-layer="3494" from-port="0" to-layer="3495" to-port="1" />
+		<edge from-layer="3495" from-port="2" to-layer="3497" to-port="0" />
+		<edge from-layer="3496" from-port="0" to-layer="3497" to-port="1" />
+		<edge from-layer="3497" from-port="2" to-layer="3506" to-port="0" />
+		<edge from-layer="3498" from-port="0" to-layer="3499" to-port="0" />
+		<edge from-layer="3499" from-port="1" to-layer="3502" to-port="0" />
+		<edge from-layer="3500" from-port="0" to-layer="3501" to-port="0" />
+		<edge from-layer="3501" from-port="1" to-layer="3502" to-port="1" />
+		<edge from-layer="3502" from-port="2" to-layer="3504" to-port="0" />
+		<edge from-layer="3503" from-port="0" to-layer="3504" to-port="1" />
+		<edge from-layer="3504" from-port="2" to-layer="3505" to-port="0" />
+		<edge from-layer="3505" from-port="1" to-layer="3506" to-port="1" />
+		<edge from-layer="3506" from-port="2" to-layer="3508" to-port="0" />
+		<edge from-layer="3507" from-port="0" to-layer="3508" to-port="1" />
+		<edge from-layer="3508" from-port="2" to-layer="3510" to-port="0" />
+		<edge from-layer="3509" from-port="0" to-layer="3510" to-port="1" />
+		<edge from-layer="3510" from-port="2" to-layer="3514" to-port="0" />
+		<edge from-layer="3510" from-port="2" to-layer="3552" to-port="0" />
+		<edge from-layer="3510" from-port="2" to-layer="3591" to-port="0" />
+		<edge from-layer="3511" from-port="0" to-layer="3514" to-port="1" />
+		<edge from-layer="3512" from-port="0" to-layer="3514" to-port="2" />
+		<edge from-layer="3513" from-port="0" to-layer="3514" to-port="3" />
+		<edge from-layer="3514" from-port="5" to-layer="3516" to-port="0" />
+		<edge from-layer="3515" from-port="0" to-layer="3516" to-port="1" />
+		<edge from-layer="3516" from-port="2" to-layer="3520" to-port="0" />
+		<edge from-layer="3516" from-port="2" to-layer="3538" to-port="0" />
+		<edge from-layer="3517" from-port="0" to-layer="3520" to-port="1" />
+		<edge from-layer="3518" from-port="0" to-layer="3520" to-port="2" />
+		<edge from-layer="3519" from-port="0" to-layer="3520" to-port="3" />
+		<edge from-layer="3520" from-port="5" to-layer="3521" to-port="0" />
+		<edge from-layer="3520" from-port="5" to-layer="3525" to-port="0" />
+		<edge from-layer="3520" from-port="5" to-layer="3531" to-port="0" />
+		<edge from-layer="3521" from-port="2" to-layer="3534" to-port="0" />
+		<edge from-layer="3522" from-port="0" to-layer="3525" to-port="1" />
+		<edge from-layer="3523" from-port="0" to-layer="3525" to-port="2" />
+		<edge from-layer="3524" from-port="0" to-layer="3525" to-port="3" />
+		<edge from-layer="3525" from-port="5" to-layer="3527" to-port="0" />
+		<edge from-layer="3526" from-port="0" to-layer="3527" to-port="1" />
+		<edge from-layer="3527" from-port="2" to-layer="3532" to-port="0" />
+		<edge from-layer="3528" from-port="0" to-layer="3531" to-port="1" />
+		<edge from-layer="3529" from-port="0" to-layer="3531" to-port="2" />
+		<edge from-layer="3530" from-port="0" to-layer="3531" to-port="3" />
+		<edge from-layer="3531" from-port="5" to-layer="3532" to-port="1" />
+		<edge from-layer="3532" from-port="2" to-layer="3533" to-port="0" />
+		<edge from-layer="3533" from-port="2" to-layer="3534" to-port="1" />
+		<edge from-layer="3534" from-port="2" to-layer="3539" to-port="0" />
+		<edge from-layer="3535" from-port="0" to-layer="3538" to-port="1" />
+		<edge from-layer="3536" from-port="0" to-layer="3538" to-port="2" />
+		<edge from-layer="3537" from-port="0" to-layer="3538" to-port="3" />
+		<edge from-layer="3538" from-port="5" to-layer="3539" to-port="1" />
+		<edge from-layer="3539" from-port="2" to-layer="3603" to-port="0" />
+		<edge from-layer="3540" from-port="0" to-layer="3545" to-port="0" />
+		<edge from-layer="3541" from-port="0" to-layer="3544" to-port="1" />
+		<edge from-layer="3542" from-port="0" to-layer="3544" to-port="2" />
+		<edge from-layer="3543" from-port="0" to-layer="3544" to-port="3" />
+		<edge from-layer="3544" from-port="4" to-layer="3545" to-port="1" />
+		<edge from-layer="3545" from-port="2" to-layer="3546" to-port="0" />
+		<edge from-layer="3546" from-port="1" to-layer="3548" to-port="0" />
+		<edge from-layer="3547" from-port="0" to-layer="3548" to-port="2" />
+		<edge from-layer="3548" from-port="3" to-layer="3596" to-port="0" />
+		<edge from-layer="3548" from-port="3" to-layer="3578" to-port="0" />
+		<edge from-layer="3549" from-port="0" to-layer="3552" to-port="1" />
+		<edge from-layer="3550" from-port="0" to-layer="3552" to-port="2" />
+		<edge from-layer="3551" from-port="0" to-layer="3552" to-port="3" />
+		<edge from-layer="3552" from-port="5" to-layer="3554" to-port="0" />
+		<edge from-layer="3553" from-port="0" to-layer="3554" to-port="1" />
+		<edge from-layer="3554" from-port="2" to-layer="3576" to-port="0" />
+		<edge from-layer="3554" from-port="2" to-layer="3558" to-port="0" />
+		<edge from-layer="3555" from-port="0" to-layer="3558" to-port="1" />
+		<edge from-layer="3556" from-port="0" to-layer="3558" to-port="2" />
+		<edge from-layer="3557" from-port="0" to-layer="3558" to-port="3" />
+		<edge from-layer="3558" from-port="5" to-layer="3569" to-port="0" />
+		<edge from-layer="3558" from-port="5" to-layer="3563" to-port="0" />
+		<edge from-layer="3558" from-port="5" to-layer="3559" to-port="0" />
+		<edge from-layer="3559" from-port="2" to-layer="3572" to-port="0" />
+		<edge from-layer="3560" from-port="0" to-layer="3563" to-port="1" />
 		<edge from-layer="3561" from-port="0" to-layer="3563" to-port="2" />
 		<edge from-layer="3562" from-port="0" to-layer="3563" to-port="3" />
-		<edge from-layer="3563" from-port="4" to-layer="3566" to-port="0" />
-		<edge from-layer="3564" from-port="0" to-layer="3565" to-port="0" />
-		<edge from-layer="3565" from-port="1" to-layer="3566" to-port="1" />
-		<edge from-layer="3566" from-port="2" to-layer="3570" to-port="0" />
-		<edge from-layer="3567" from-port="0" to-layer="3569" to-port="1" />
+		<edge from-layer="3563" from-port="5" to-layer="3565" to-port="0" />
+		<edge from-layer="3564" from-port="0" to-layer="3565" to-port="1" />
+		<edge from-layer="3565" from-port="2" to-layer="3570" to-port="0" />
+		<edge from-layer="3566" from-port="0" to-layer="3569" to-port="1" />
+		<edge from-layer="3567" from-port="0" to-layer="3569" to-port="2" />
 		<edge from-layer="3568" from-port="0" to-layer="3569" to-port="3" />
-		<edge from-layer="3569" from-port="4" to-layer="3570" to-port="1" />
-		<edge from-layer="3570" from-port="2" to-layer="3579" to-port="0" />
-		<edge from-layer="3571" from-port="0" to-layer="3572" to-port="0" />
-		<edge from-layer="3572" from-port="1" to-layer="3575" to-port="0" />
-		<edge from-layer="3573" from-port="0" to-layer="3575" to-port="1" />
-		<edge from-layer="3574" from-port="0" to-layer="3575" to-port="3" />
-		<edge from-layer="3575" from-port="4" to-layer="3577" to-port="0" />
-		<edge from-layer="3576" from-port="0" to-layer="3577" to-port="2" />
-		<edge from-layer="3577" from-port="3" to-layer="3578" to-port="0" />
-		<edge from-layer="3578" from-port="2" to-layer="3612" to-port="1" />
-		<edge from-layer="3578" from-port="2" to-layer="3579" to-port="1" />
-		<edge from-layer="3579" from-port="2" to-layer="3580" to-port="1" />
-		<edge from-layer="3580" from-port="2" to-layer="3585" to-port="0" />
-		<edge from-layer="3581" from-port="0" to-layer="3584" to-port="1" />
-		<edge from-layer="3582" from-port="0" to-layer="3584" to-port="2" />
-		<edge from-layer="3583" from-port="0" to-layer="3584" to-port="3" />
-		<edge from-layer="3584" from-port="4" to-layer="3585" to-port="1" />
-		<edge from-layer="3585" from-port="2" to-layer="3676" to-port="0" />
-		<edge from-layer="3585" from-port="2" to-layer="3643" to-port="0" />
-		<edge from-layer="3586" from-port="0" to-layer="3589" to-port="1" />
-		<edge from-layer="3587" from-port="0" to-layer="3589" to-port="2" />
-		<edge from-layer="3588" from-port="0" to-layer="3589" to-port="3" />
-		<edge from-layer="3589" from-port="4" to-layer="3590" to-port="0" />
-		<edge from-layer="3589" from-port="4" to-layer="3593" to-port="0" />
-		<edge from-layer="3589" from-port="4" to-layer="3604" to-port="0" />
-		<edge from-layer="3589" from-port="4" to-layer="3610" to-port="0" />
-		<edge from-layer="3590" from-port="2" to-layer="3613" to-port="0" />
-		<edge from-layer="3591" from-port="0" to-layer="3601" to-port="0" />
-		<edge from-layer="3592" from-port="0" to-layer="3601" to-port="1" />
-		<edge from-layer="3593" from-port="1" to-layer="3595" to-port="0" />
-		<edge from-layer="3594" from-port="0" to-layer="3595" to-port="2" />
-		<edge from-layer="3595" from-port="3" to-layer="3596" to-port="0" />
-		<edge from-layer="3596" from-port="2" to-layer="3597" to-port="0" />
-		<edge from-layer="3597" from-port="1" to-layer="3599" to-port="0" />
-		<edge from-layer="3598" from-port="0" to-layer="3599" to-port="1" />
-		<edge from-layer="3599" from-port="2" to-layer="3601" to-port="2" />
-		<edge from-layer="3600" from-port="0" to-layer="3601" to-port="3" />
-		<edge from-layer="3601" from-port="4" to-layer="3610" to-port="2" />
-		<edge from-layer="3601" from-port="4" to-layer="3604" to-port="1" />
-		<edge from-layer="3602" from-port="0" to-layer="3604" to-port="2" />
-		<edge from-layer="3603" from-port="0" to-layer="3604" to-port="3" />
-		<edge from-layer="3604" from-port="4" to-layer="3607" to-port="0" />
-		<edge from-layer="3605" from-port="0" to-layer="3606" to-port="0" />
-		<edge from-layer="3606" from-port="1" to-layer="3607" to-port="1" />
-		<edge from-layer="3607" from-port="2" to-layer="3611" to-port="0" />
-		<edge from-layer="3608" from-port="0" to-layer="3610" to-port="1" />
-		<edge from-layer="3609" from-port="0" to-layer="3610" to-port="3" />
-		<edge from-layer="3610" from-port="4" to-layer="3611" to-port="1" />
-		<edge from-layer="3611" from-port="2" to-layer="3612" to-port="0" />
-		<edge from-layer="3612" from-port="2" to-layer="3613" to-port="1" />
-		<edge from-layer="3613" from-port="2" to-layer="3618" to-port="0" />
-		<edge from-layer="3614" from-port="0" to-layer="3617" to-port="1" />
-		<edge from-layer="3615" from-port="0" to-layer="3617" to-port="2" />
-		<edge from-layer="3616" from-port="0" to-layer="3617" to-port="3" />
-		<edge from-layer="3617" from-port="4" to-layer="3618" to-port="1" />
+		<edge from-layer="3569" from-port="5" to-layer="3570" to-port="1" />
+		<edge from-layer="3570" from-port="2" to-layer="3571" to-port="0" />
+		<edge from-layer="3571" from-port="2" to-layer="3572" to-port="1" />
+		<edge from-layer="3572" from-port="2" to-layer="3577" to-port="0" />
+		<edge from-layer="3573" from-port="0" to-layer="3576" to-port="1" />
+		<edge from-layer="3574" from-port="0" to-layer="3576" to-port="2" />
+		<edge from-layer="3575" from-port="0" to-layer="3576" to-port="3" />
+		<edge from-layer="3576" from-port="5" to-layer="3577" to-port="1" />
+		<edge from-layer="3577" from-port="2" to-layer="3578" to-port="1" />
+		<edge from-layer="3578" from-port="2" to-layer="3603" to-port="1" />
+		<edge from-layer="3578" from-port="2" to-layer="5131" to-port="0" />
+		<edge from-layer="3579" from-port="0" to-layer="3584" to-port="0" />
+		<edge from-layer="3580" from-port="0" to-layer="3583" to-port="1" />
+		<edge from-layer="3581" from-port="0" to-layer="3583" to-port="2" />
+		<edge from-layer="3582" from-port="0" to-layer="3583" to-port="3" />
+		<edge from-layer="3583" from-port="4" to-layer="3584" to-port="1" />
+		<edge from-layer="3584" from-port="2" to-layer="3585" to-port="0" />
+		<edge from-layer="3585" from-port="1" to-layer="3587" to-port="0" />
+		<edge from-layer="3586" from-port="0" to-layer="3587" to-port="2" />
+		<edge from-layer="3587" from-port="3" to-layer="3594" to-port="0" />
+		<edge from-layer="3588" from-port="0" to-layer="3591" to-port="1" />
+		<edge from-layer="3589" from-port="0" to-layer="3591" to-port="2" />
+		<edge from-layer="3590" from-port="0" to-layer="3591" to-port="3" />
+		<edge from-layer="3591" from-port="5" to-layer="3593" to-port="0" />
+		<edge from-layer="3592" from-port="0" to-layer="3593" to-port="1" />
+		<edge from-layer="3593" from-port="2" to-layer="3594" to-port="1" />
+		<edge from-layer="3594" from-port="2" to-layer="3603" to-port="2" />
+		<edge from-layer="3594" from-port="2" to-layer="5130" to-port="0" />
+		<edge from-layer="3595" from-port="0" to-layer="3602" to-port="1" />
+		<edge from-layer="3596" from-port="1" to-layer="3599" to-port="0" />
+		<edge from-layer="3597" from-port="0" to-layer="3599" to-port="1" />
+		<edge from-layer="3598" from-port="0" to-layer="3599" to-port="2" />
+		<edge from-layer="3599" from-port="3" to-layer="3600" to-port="0" />
+		<edge from-layer="3600" from-port="2" to-layer="3602" to-port="2" />
+		<edge from-layer="3601" from-port="0" to-layer="3602" to-port="3" />
+		<edge from-layer="3602" from-port="5" to-layer="3603" to-port="3" />
+		<edge from-layer="3603" from-port="4" to-layer="3605" to-port="0" />
+		<edge from-layer="3604" from-port="0" to-layer="3605" to-port="1" />
+		<edge from-layer="3605" from-port="2" to-layer="3607" to-port="0" />
+		<edge from-layer="3606" from-port="0" to-layer="3607" to-port="1" />
+		<edge from-layer="3607" from-port="2" to-layer="3616" to-port="0" />
+		<edge from-layer="3608" from-port="0" to-layer="3609" to-port="0" />
+		<edge from-layer="3609" from-port="1" to-layer="3612" to-port="0" />
+		<edge from-layer="3610" from-port="0" to-layer="3611" to-port="0" />
+		<edge from-layer="3611" from-port="1" to-layer="3612" to-port="1" />
+		<edge from-layer="3612" from-port="2" to-layer="3614" to-port="0" />
+		<edge from-layer="3613" from-port="0" to-layer="3614" to-port="1" />
+		<edge from-layer="3614" from-port="2" to-layer="3615" to-port="0" />
+		<edge from-layer="3615" from-port="1" to-layer="3616" to-port="1" />
+		<edge from-layer="3616" from-port="2" to-layer="3618" to-port="0" />
+		<edge from-layer="3617" from-port="0" to-layer="3618" to-port="1" />
 		<edge from-layer="3618" from-port="2" to-layer="3619" to-port="1" />
-		<edge from-layer="3619" from-port="2" to-layer="3640" to-port="0" />
-		<edge from-layer="3619" from-port="2" to-layer="3676" to-port="1" />
-		<edge from-layer="3619" from-port="2" to-layer="7710" to-port="0" />
-		<edge from-layer="3620" from-port="0" to-layer="3621" to-port="0" />
-		<edge from-layer="3621" from-port="1" to-layer="3626" to-port="0" />
-		<edge from-layer="3622" from-port="0" to-layer="3625" to-port="1" />
-		<edge from-layer="3623" from-port="0" to-layer="3625" to-port="2" />
-		<edge from-layer="3624" from-port="0" to-layer="3625" to-port="3" />
-		<edge from-layer="3625" from-port="4" to-layer="3626" to-port="1" />
-		<edge from-layer="3626" from-port="2" to-layer="3627" to-port="0" />
-		<edge from-layer="3627" from-port="1" to-layer="3629" to-port="0" />
-		<edge from-layer="3628" from-port="0" to-layer="3629" to-port="2" />
-		<edge from-layer="3629" from-port="3" to-layer="3636" to-port="0" />
-		<edge from-layer="3630" from-port="0" to-layer="3633" to-port="1" />
-		<edge from-layer="3631" from-port="0" to-layer="3633" to-port="2" />
-		<edge from-layer="3632" from-port="0" to-layer="3633" to-port="3" />
-		<edge from-layer="3633" from-port="4" to-layer="3635" to-port="0" />
-		<edge from-layer="3634" from-port="0" to-layer="3635" to-port="1" />
-		<edge from-layer="3635" from-port="2" to-layer="3636" to-port="1" />
-		<edge from-layer="3636" from-port="2" to-layer="3676" to-port="2" />
-		<edge from-layer="3636" from-port="2" to-layer="7709" to-port="0" />
-		<edge from-layer="3637" from-port="0" to-layer="3655" to-port="0" />
-		<edge from-layer="3638" from-port="0" to-layer="3650" to-port="0" />
-		<edge from-layer="3639" from-port="0" to-layer="3653" to-port="1" />
-		<edge from-layer="3639" from-port="0" to-layer="3650" to-port="1" />
-		<edge from-layer="3640" from-port="1" to-layer="3642" to-port="0" />
-		<edge from-layer="3641" from-port="0" to-layer="3642" to-port="2" />
-		<edge from-layer="3642" from-port="3" to-layer="3646" to-port="0" />
-		<edge from-layer="3642" from-port="3" to-layer="3652" to-port="0" />
-		<edge from-layer="3643" from-port="1" to-layer="3645" to-port="0" />
-		<edge from-layer="3643" from-port="1" to-layer="3671" to-port="0" />
-		<edge from-layer="3644" from-port="0" to-layer="3645" to-port="2" />
-		<edge from-layer="3645" from-port="3" to-layer="3646" to-port="1" />
-		<edge from-layer="3646" from-port="2" to-layer="3648" to-port="0" />
-		<edge from-layer="3647" from-port="0" to-layer="3648" to-port="1" />
-		<edge from-layer="3647" from-port="0" to-layer="3652" to-port="1" />
-		<edge from-layer="3648" from-port="2" to-layer="3650" to-port="2" />
-		<edge from-layer="3649" from-port="0" to-layer="3653" to-port="3" />
-		<edge from-layer="3649" from-port="0" to-layer="3650" to-port="3" />
-		<edge from-layer="3650" from-port="4" to-layer="3655" to-port="1" />
-		<edge from-layer="3651" from-port="0" to-layer="3653" to-port="0" />
-		<edge from-layer="3652" from-port="2" to-layer="3653" to-port="2" />
-		<edge from-layer="3652" from-port="2" to-layer="3660" to-port="2" />
-		<edge from-layer="3653" from-port="4" to-layer="3655" to-port="2" />
-		<edge from-layer="3654" from-port="0" to-layer="3655" to-port="3" />
-		<edge from-layer="3655" from-port="4" to-layer="3662" to-port="0" />
-		<edge from-layer="3656" from-port="0" to-layer="3662" to-port="1" />
-		<edge from-layer="3657" from-port="0" to-layer="3660" to-port="0" />
-		<edge from-layer="3658" from-port="0" to-layer="3660" to-port="1" />
-		<edge from-layer="3659" from-port="0" to-layer="3660" to-port="3" />
-		<edge from-layer="3660" from-port="4" to-layer="3662" to-port="2" />
-		<edge from-layer="3661" from-port="0" to-layer="3662" to-port="3" />
-		<edge from-layer="3662" from-port="4" to-layer="3666" to-port="0" />
-		<edge from-layer="3663" from-port="0" to-layer="3664" to-port="0" />
-		<edge from-layer="3664" from-port="1" to-layer="3666" to-port="1" />
-		<edge from-layer="3665" from-port="0" to-layer="3666" to-port="2" />
-		<edge from-layer="3666" from-port="3" to-layer="3667" to-port="0" />
-		<edge from-layer="3667" from-port="2" to-layer="3668" to-port="0" />
-		<edge from-layer="3667" from-port="2" to-layer="3675" to-port="0" />
-		<edge from-layer="3668" from-port="1" to-layer="3674" to-port="0" />
-		<edge from-layer="3669" from-port="0" to-layer="3671" to-port="1" />
-		<edge from-layer="3670" from-port="0" to-layer="3671" to-port="2" />
-		<edge from-layer="3671" from-port="3" to-layer="3672" to-port="0" />
-		<edge from-layer="3672" from-port="4" to-layer="3673" to-port="0" />
-		<edge from-layer="3673" from-port="1" to-layer="3674" to-port="1" />
-		<edge from-layer="3674" from-port="2" to-layer="3675" to-port="1" />
-		<edge from-layer="3675" from-port="2" to-layer="3676" to-port="3" />
-		<edge from-layer="3676" from-port="4" to-layer="3678" to-port="0" />
-		<edge from-layer="3677" from-port="0" to-layer="3678" to-port="1" />
-		<edge from-layer="3678" from-port="2" to-layer="3680" to-port="0" />
-		<edge from-layer="3679" from-port="0" to-layer="3680" to-port="1" />
-		<edge from-layer="3680" from-port="2" to-layer="3689" to-port="0" />
-		<edge from-layer="3681" from-port="0" to-layer="3682" to-port="0" />
-		<edge from-layer="3682" from-port="1" to-layer="3685" to-port="0" />
-		<edge from-layer="3683" from-port="0" to-layer="3684" to-port="0" />
-		<edge from-layer="3684" from-port="1" to-layer="3685" to-port="1" />
-		<edge from-layer="3685" from-port="2" to-layer="3687" to-port="0" />
-		<edge from-layer="3686" from-port="0" to-layer="3687" to-port="1" />
-		<edge from-layer="3687" from-port="2" to-layer="3688" to-port="0" />
-		<edge from-layer="3688" from-port="1" to-layer="3689" to-port="1" />
-		<edge from-layer="3689" from-port="2" to-layer="3692" to-port="0" />
-		<edge from-layer="3690" from-port="0" to-layer="3691" to-port="0" />
-		<edge from-layer="3691" from-port="1" to-layer="3692" to-port="1" />
-		<edge from-layer="3692" from-port="2" to-layer="3693" to-port="1" />
-		<edge from-layer="3693" from-port="2" to-layer="3694" to-port="0" />
-		<edge from-layer="3694" from-port="2" to-layer="3936" to-port="1" />
-		<edge from-layer="3694" from-port="2" to-layer="3696" to-port="0" />
-		<edge from-layer="3695" from-port="0" to-layer="3696" to-port="1" />
-		<edge from-layer="3696" from-port="2" to-layer="3730" to-port="0" />
-		<edge from-layer="3696" from-port="2" to-layer="3699" to-port="0" />
-		<edge from-layer="3697" from-port="0" to-layer="3698" to-port="0" />
-		<edge from-layer="3698" from-port="1" to-layer="3699" to-port="1" />
-		<edge from-layer="3699" from-port="2" to-layer="3702" to-port="0" />
-		<edge from-layer="3700" from-port="0" to-layer="3701" to-port="0" />
-		<edge from-layer="3701" from-port="1" to-layer="3702" to-port="1" />
-		<edge from-layer="3702" from-port="2" to-layer="3711" to-port="0" />
-		<edge from-layer="3703" from-port="0" to-layer="3704" to-port="0" />
-		<edge from-layer="3704" from-port="1" to-layer="3707" to-port="0" />
-		<edge from-layer="3705" from-port="0" to-layer="3706" to-port="0" />
-		<edge from-layer="3706" from-port="1" to-layer="3707" to-port="1" />
-		<edge from-layer="3707" from-port="2" to-layer="3709" to-port="0" />
-		<edge from-layer="3708" from-port="0" to-layer="3709" to-port="1" />
-		<edge from-layer="3709" from-port="2" to-layer="3710" to-port="0" />
-		<edge from-layer="3710" from-port="1" to-layer="3711" to-port="1" />
-		<edge from-layer="3711" from-port="2" to-layer="3714" to-port="0" />
-		<edge from-layer="3712" from-port="0" to-layer="3713" to-port="0" />
-		<edge from-layer="3713" from-port="1" to-layer="3714" to-port="1" />
-		<edge from-layer="3714" from-port="2" to-layer="3715" to-port="0" />
-		<edge from-layer="3715" from-port="1" to-layer="3724" to-port="0" />
-		<edge from-layer="3716" from-port="0" to-layer="3717" to-port="0" />
-		<edge from-layer="3717" from-port="1" to-layer="3720" to-port="0" />
-		<edge from-layer="3718" from-port="0" to-layer="3719" to-port="0" />
-		<edge from-layer="3719" from-port="1" to-layer="3720" to-port="1" />
-		<edge from-layer="3720" from-port="2" to-layer="3722" to-port="0" />
-		<edge from-layer="3721" from-port="0" to-layer="3722" to-port="1" />
-		<edge from-layer="3722" from-port="2" to-layer="3723" to-port="0" />
-		<edge from-layer="3723" from-port="1" to-layer="3724" to-port="1" />
-		<edge from-layer="3724" from-port="2" to-layer="3727" to-port="0" />
-		<edge from-layer="3725" from-port="0" to-layer="3726" to-port="0" />
-		<edge from-layer="3726" from-port="1" to-layer="3727" to-port="1" />
-		<edge from-layer="3727" from-port="2" to-layer="3935" to-port="0" />
-		<edge from-layer="3728" from-port="0" to-layer="3729" to-port="0" />
-		<edge from-layer="3729" from-port="1" to-layer="3730" to-port="1" />
-		<edge from-layer="3730" from-port="2" to-layer="3733" to-port="0" />
-		<edge from-layer="3731" from-port="0" to-layer="3732" to-port="0" />
-		<edge from-layer="3732" from-port="1" to-layer="3733" to-port="1" />
-		<edge from-layer="3733" from-port="2" to-layer="3742" to-port="0" />
-		<edge from-layer="3734" from-port="0" to-layer="3735" to-port="0" />
-		<edge from-layer="3735" from-port="1" to-layer="3738" to-port="0" />
-		<edge from-layer="3736" from-port="0" to-layer="3737" to-port="0" />
-		<edge from-layer="3737" from-port="1" to-layer="3738" to-port="1" />
-		<edge from-layer="3738" from-port="2" to-layer="3740" to-port="0" />
-		<edge from-layer="3739" from-port="0" to-layer="3740" to-port="1" />
-		<edge from-layer="3740" from-port="2" to-layer="3741" to-port="0" />
-		<edge from-layer="3741" from-port="1" to-layer="3742" to-port="1" />
-		<edge from-layer="3742" from-port="2" to-layer="3745" to-port="0" />
-		<edge from-layer="3743" from-port="0" to-layer="3744" to-port="0" />
-		<edge from-layer="3744" from-port="1" to-layer="3745" to-port="1" />
-		<edge from-layer="3745" from-port="2" to-layer="3747" to-port="0" />
-		<edge from-layer="3746" from-port="0" to-layer="3747" to-port="1" />
-		<edge from-layer="3747" from-port="2" to-layer="3875" to-port="0" />
-		<edge from-layer="3747" from-port="2" to-layer="3751" to-port="0" />
-		<edge from-layer="3747" from-port="2" to-layer="3764" to-port="0" />
-		<edge from-layer="3748" from-port="0" to-layer="3751" to-port="1" />
-		<edge from-layer="3749" from-port="0" to-layer="3751" to-port="2" />
-		<edge from-layer="3750" from-port="0" to-layer="3751" to-port="3" />
-		<edge from-layer="3751" from-port="4" to-layer="3753" to-port="0" />
-		<edge from-layer="3752" from-port="0" to-layer="3753" to-port="1" />
-		<edge from-layer="3753" from-port="2" to-layer="3826" to-port="0" />
-		<edge from-layer="3753" from-port="2" to-layer="3757" to-port="0" />
-		<edge from-layer="3754" from-port="0" to-layer="3757" to-port="1" />
-		<edge from-layer="3755" from-port="0" to-layer="3757" to-port="2" />
-		<edge from-layer="3756" from-port="0" to-layer="3757" to-port="3" />
-		<edge from-layer="3757" from-port="4" to-layer="3805" to-port="0" />
-		<edge from-layer="3757" from-port="4" to-layer="3811" to-port="0" />
-		<edge from-layer="3757" from-port="4" to-layer="3794" to-port="0" />
-		<edge from-layer="3757" from-port="4" to-layer="3791" to-port="0" />
-		<edge from-layer="3758" from-port="0" to-layer="3759" to-port="0" />
-		<edge from-layer="3759" from-port="1" to-layer="3787" to-port="0" />
-		<edge from-layer="3760" from-port="0" to-layer="3787" to-port="1" />
-		<edge from-layer="3761" from-port="0" to-layer="3764" to-port="1" />
-		<edge from-layer="3762" from-port="0" to-layer="3764" to-port="2" />
-		<edge from-layer="3763" from-port="0" to-layer="3764" to-port="3" />
-		<edge from-layer="3764" from-port="4" to-layer="3766" to-port="0" />
-		<edge from-layer="3765" from-port="0" to-layer="3766" to-port="1" />
-		<edge from-layer="3766" from-port="2" to-layer="3859" to-port="0" />
-		<edge from-layer="3766" from-port="2" to-layer="3767" to-port="0" />
-		<edge from-layer="3766" from-port="2" to-layer="3831" to-port="0" />
-		<edge from-layer="3767" from-port="1" to-layer="3769" to-port="0" />
-		<edge from-layer="3768" from-port="0" to-layer="3769" to-port="2" />
-		<edge from-layer="3769" from-port="3" to-layer="3783" to-port="0" />
-		<edge from-layer="3770" from-port="0" to-layer="3771" to-port="0" />
-		<edge from-layer="3771" from-port="1" to-layer="3776" to-port="0" />
-		<edge from-layer="3772" from-port="0" to-layer="3775" to-port="1" />
-		<edge from-layer="3773" from-port="0" to-layer="3775" to-port="2" />
-		<edge from-layer="3774" from-port="0" to-layer="3775" to-port="3" />
-		<edge from-layer="3775" from-port="4" to-layer="3776" to-port="1" />
-		<edge from-layer="3776" from-port="2" to-layer="3777" to-port="0" />
-		<edge from-layer="3777" from-port="1" to-layer="3779" to-port="0" />
-		<edge from-layer="3778" from-port="0" to-layer="3779" to-port="2" />
-		<edge from-layer="3779" from-port="3" to-layer="3780" to-port="0" />
-		<edge from-layer="3779" from-port="3" to-layer="3861" to-port="0" />
-		<edge from-layer="3780" from-port="1" to-layer="3782" to-port="0" />
-		<edge from-layer="3781" from-port="0" to-layer="3782" to-port="2" />
-		<edge from-layer="3782" from-port="3" to-layer="3783" to-port="1" />
-		<edge from-layer="3783" from-port="2" to-layer="3785" to-port="0" />
-		<edge from-layer="3784" from-port="0" to-layer="3785" to-port="1" />
-		<edge from-layer="3785" from-port="2" to-layer="3817" to-port="2" />
-		<edge from-layer="3785" from-port="2" to-layer="3787" to-port="2" />
-		<edge from-layer="3786" from-port="0" to-layer="3787" to-port="3" />
-		<edge from-layer="3787" from-port="4" to-layer="3789" to-port="0" />
-		<edge from-layer="3788" from-port="0" to-layer="3789" to-port="2" />
-		<edge from-layer="3789" from-port="3" to-layer="3790" to-port="0" />
-		<edge from-layer="3790" from-port="2" to-layer="3791" to-port="1" />
-		<edge from-layer="3790" from-port="2" to-layer="3832" to-port="1" />
-		<edge from-layer="3791" from-port="2" to-layer="3822" to-port="0" />
-		<edge from-layer="3792" from-port="0" to-layer="3802" to-port="0" />
-		<edge from-layer="3793" from-port="0" to-layer="3802" to-port="1" />
-		<edge from-layer="3794" from-port="1" to-layer="3796" to-port="0" />
-		<edge from-layer="3795" from-port="0" to-layer="3796" to-port="2" />
-		<edge from-layer="3796" from-port="3" to-layer="3797" to-port="0" />
-		<edge from-layer="3797" from-port="2" to-layer="3798" to-port="0" />
-		<edge from-layer="3798" from-port="1" to-layer="3800" to-port="0" />
-		<edge from-layer="3799" from-port="0" to-layer="3800" to-port="1" />
-		<edge from-layer="3800" from-port="2" to-layer="3802" to-port="2" />
-		<edge from-layer="3801" from-port="0" to-layer="3802" to-port="3" />
-		<edge from-layer="3802" from-port="4" to-layer="3811" to-port="2" />
-		<edge from-layer="3802" from-port="4" to-layer="3805" to-port="1" />
-		<edge from-layer="3803" from-port="0" to-layer="3805" to-port="2" />
-		<edge from-layer="3804" from-port="0" to-layer="3805" to-port="3" />
-		<edge from-layer="3805" from-port="4" to-layer="3808" to-port="0" />
-		<edge from-layer="3806" from-port="0" to-layer="3807" to-port="0" />
-		<edge from-layer="3807" from-port="1" to-layer="3808" to-port="1" />
-		<edge from-layer="3808" from-port="2" to-layer="3812" to-port="0" />
-		<edge from-layer="3809" from-port="0" to-layer="3811" to-port="1" />
-		<edge from-layer="3810" from-port="0" to-layer="3811" to-port="3" />
-		<edge from-layer="3811" from-port="4" to-layer="3812" to-port="1" />
-		<edge from-layer="3812" from-port="2" to-layer="3821" to-port="0" />
-		<edge from-layer="3813" from-port="0" to-layer="3814" to-port="0" />
-		<edge from-layer="3814" from-port="1" to-layer="3817" to-port="0" />
-		<edge from-layer="3815" from-port="0" to-layer="3817" to-port="1" />
-		<edge from-layer="3816" from-port="0" to-layer="3817" to-port="3" />
-		<edge from-layer="3817" from-port="4" to-layer="3819" to-port="0" />
-		<edge from-layer="3818" from-port="0" to-layer="3819" to-port="2" />
-		<edge from-layer="3819" from-port="3" to-layer="3820" to-port="0" />
-		<edge from-layer="3820" from-port="2" to-layer="3854" to-port="1" />
-		<edge from-layer="3820" from-port="2" to-layer="3821" to-port="1" />
-		<edge from-layer="3821" from-port="2" to-layer="3822" to-port="1" />
-		<edge from-layer="3822" from-port="2" to-layer="3827" to-port="0" />
-		<edge from-layer="3823" from-port="0" to-layer="3826" to-port="1" />
-		<edge from-layer="3824" from-port="0" to-layer="3826" to-port="2" />
-		<edge from-layer="3825" from-port="0" to-layer="3826" to-port="3" />
-		<edge from-layer="3826" from-port="4" to-layer="3827" to-port="1" />
-		<edge from-layer="3827" from-port="2" to-layer="3918" to-port="0" />
-		<edge from-layer="3827" from-port="2" to-layer="3885" to-port="0" />
-		<edge from-layer="3828" from-port="0" to-layer="3831" to-port="1" />
-		<edge from-layer="3829" from-port="0" to-layer="3831" to-port="2" />
-		<edge from-layer="3830" from-port="0" to-layer="3831" to-port="3" />
-		<edge from-layer="3831" from-port="4" to-layer="3852" to-port="0" />
-		<edge from-layer="3831" from-port="4" to-layer="3835" to-port="0" />
-		<edge from-layer="3831" from-port="4" to-layer="3832" to-port="0" />
-		<edge from-layer="3831" from-port="4" to-layer="3846" to-port="0" />
-		<edge from-layer="3832" from-port="2" to-layer="3855" to-port="0" />
-		<edge from-layer="3833" from-port="0" to-layer="3843" to-port="0" />
-		<edge from-layer="3834" from-port="0" to-layer="3843" to-port="1" />
-		<edge from-layer="3835" from-port="1" to-layer="3837" to-port="0" />
-		<edge from-layer="3836" from-port="0" to-layer="3837" to-port="2" />
-		<edge from-layer="3837" from-port="3" to-layer="3838" to-port="0" />
-		<edge from-layer="3838" from-port="2" to-layer="3839" to-port="0" />
-		<edge from-layer="3839" from-port="1" to-layer="3841" to-port="0" />
-		<edge from-layer="3840" from-port="0" to-layer="3841" to-port="1" />
-		<edge from-layer="3841" from-port="2" to-layer="3843" to-port="2" />
-		<edge from-layer="3842" from-port="0" to-layer="3843" to-port="3" />
-		<edge from-layer="3843" from-port="4" to-layer="3852" to-port="2" />
-		<edge from-layer="3843" from-port="4" to-layer="3846" to-port="1" />
-		<edge from-layer="3844" from-port="0" to-layer="3846" to-port="2" />
-		<edge from-layer="3845" from-port="0" to-layer="3846" to-port="3" />
-		<edge from-layer="3846" from-port="4" to-layer="3849" to-port="0" />
-		<edge from-layer="3847" from-port="0" to-layer="3848" to-port="0" />
-		<edge from-layer="3848" from-port="1" to-layer="3849" to-port="1" />
-		<edge from-layer="3849" from-port="2" to-layer="3853" to-port="0" />
-		<edge from-layer="3850" from-port="0" to-layer="3852" to-port="1" />
-		<edge from-layer="3851" from-port="0" to-layer="3852" to-port="3" />
-		<edge from-layer="3852" from-port="4" to-layer="3853" to-port="1" />
-		<edge from-layer="3853" from-port="2" to-layer="3854" to-port="0" />
-		<edge from-layer="3854" from-port="2" to-layer="3855" to-port="1" />
-		<edge from-layer="3855" from-port="2" to-layer="3860" to-port="0" />
-		<edge from-layer="3856" from-port="0" to-layer="3859" to-port="1" />
-		<edge from-layer="3857" from-port="0" to-layer="3859" to-port="2" />
-		<edge from-layer="3858" from-port="0" to-layer="3859" to-port="3" />
-		<edge from-layer="3859" from-port="4" to-layer="3860" to-port="1" />
-		<edge from-layer="3860" from-port="2" to-layer="3861" to-port="1" />
-		<edge from-layer="3861" from-port="2" to-layer="3918" to-port="1" />
-		<edge from-layer="3861" from-port="2" to-layer="3882" to-port="0" />
-		<edge from-layer="3861" from-port="2" to-layer="7708" to-port="0" />
-		<edge from-layer="3862" from-port="0" to-layer="3863" to-port="0" />
-		<edge from-layer="3863" from-port="1" to-layer="3868" to-port="0" />
-		<edge from-layer="3864" from-port="0" to-layer="3867" to-port="1" />
-		<edge from-layer="3865" from-port="0" to-layer="3867" to-port="2" />
-		<edge from-layer="3866" from-port="0" to-layer="3867" to-port="3" />
-		<edge from-layer="3867" from-port="4" to-layer="3868" to-port="1" />
-		<edge from-layer="3868" from-port="2" to-layer="3869" to-port="0" />
-		<edge from-layer="3869" from-port="1" to-layer="3871" to-port="0" />
-		<edge from-layer="3870" from-port="0" to-layer="3871" to-port="2" />
-		<edge from-layer="3871" from-port="3" to-layer="3878" to-port="0" />
-		<edge from-layer="3872" from-port="0" to-layer="3875" to-port="1" />
-		<edge from-layer="3873" from-port="0" to-layer="3875" to-port="2" />
-		<edge from-layer="3874" from-port="0" to-layer="3875" to-port="3" />
-		<edge from-layer="3875" from-port="4" to-layer="3877" to-port="0" />
-		<edge from-layer="3876" from-port="0" to-layer="3877" to-port="1" />
-		<edge from-layer="3877" from-port="2" to-layer="3878" to-port="1" />
-		<edge from-layer="3878" from-port="2" to-layer="7707" to-port="0" />
-		<edge from-layer="3878" from-port="2" to-layer="3918" to-port="2" />
-		<edge from-layer="3879" from-port="0" to-layer="3897" to-port="0" />
-		<edge from-layer="3880" from-port="0" to-layer="3892" to-port="0" />
-		<edge from-layer="3881" from-port="0" to-layer="3892" to-port="1" />
-		<edge from-layer="3881" from-port="0" to-layer="3895" to-port="1" />
-		<edge from-layer="3882" from-port="1" to-layer="3884" to-port="0" />
-		<edge from-layer="3883" from-port="0" to-layer="3884" to-port="2" />
-		<edge from-layer="3884" from-port="3" to-layer="3888" to-port="0" />
-		<edge from-layer="3884" from-port="3" to-layer="3894" to-port="0" />
-		<edge from-layer="3885" from-port="1" to-layer="3887" to-port="0" />
-		<edge from-layer="3885" from-port="1" to-layer="3913" to-port="0" />
-		<edge from-layer="3886" from-port="0" to-layer="3887" to-port="2" />
-		<edge from-layer="3887" from-port="3" to-layer="3888" to-port="1" />
-		<edge from-layer="3888" from-port="2" to-layer="3890" to-port="0" />
-		<edge from-layer="3889" from-port="0" to-layer="3890" to-port="1" />
-		<edge from-layer="3889" from-port="0" to-layer="3894" to-port="1" />
-		<edge from-layer="3890" from-port="2" to-layer="3892" to-port="2" />
-		<edge from-layer="3891" from-port="0" to-layer="3895" to-port="3" />
+		<edge from-layer="3619" from-port="2" to-layer="3620" to-port="0" />
+		<edge from-layer="3620" from-port="2" to-layer="3622" to-port="0" />
+		<edge from-layer="3620" from-port="2" to-layer="3778" to-port="1" />
+		<edge from-layer="3621" from-port="0" to-layer="3622" to-port="1" />
+		<edge from-layer="3622" from-port="2" to-layer="3653" to-port="0" />
+		<edge from-layer="3622" from-port="2" to-layer="3624" to-port="0" />
+		<edge from-layer="3623" from-port="0" to-layer="3624" to-port="1" />
+		<edge from-layer="3624" from-port="2" to-layer="3626" to-port="0" />
+		<edge from-layer="3625" from-port="0" to-layer="3626" to-port="1" />
+		<edge from-layer="3626" from-port="2" to-layer="3635" to-port="0" />
+		<edge from-layer="3627" from-port="0" to-layer="3628" to-port="0" />
+		<edge from-layer="3628" from-port="1" to-layer="3631" to-port="0" />
+		<edge from-layer="3629" from-port="0" to-layer="3630" to-port="0" />
+		<edge from-layer="3630" from-port="1" to-layer="3631" to-port="1" />
+		<edge from-layer="3631" from-port="2" to-layer="3633" to-port="0" />
+		<edge from-layer="3632" from-port="0" to-layer="3633" to-port="1" />
+		<edge from-layer="3633" from-port="2" to-layer="3634" to-port="0" />
+		<edge from-layer="3634" from-port="1" to-layer="3635" to-port="1" />
+		<edge from-layer="3635" from-port="2" to-layer="3638" to-port="0" />
+		<edge from-layer="3636" from-port="0" to-layer="3637" to-port="0" />
+		<edge from-layer="3637" from-port="1" to-layer="3638" to-port="1" />
+		<edge from-layer="3638" from-port="2" to-layer="3639" to-port="0" />
+		<edge from-layer="3639" from-port="1" to-layer="3648" to-port="0" />
+		<edge from-layer="3640" from-port="0" to-layer="3641" to-port="0" />
+		<edge from-layer="3641" from-port="1" to-layer="3644" to-port="0" />
+		<edge from-layer="3642" from-port="0" to-layer="3643" to-port="0" />
+		<edge from-layer="3643" from-port="1" to-layer="3644" to-port="1" />
+		<edge from-layer="3644" from-port="2" to-layer="3646" to-port="0" />
+		<edge from-layer="3645" from-port="0" to-layer="3646" to-port="1" />
+		<edge from-layer="3646" from-port="2" to-layer="3647" to-port="0" />
+		<edge from-layer="3647" from-port="1" to-layer="3648" to-port="1" />
+		<edge from-layer="3648" from-port="2" to-layer="3651" to-port="0" />
+		<edge from-layer="3649" from-port="0" to-layer="3650" to-port="0" />
+		<edge from-layer="3650" from-port="1" to-layer="3651" to-port="1" />
+		<edge from-layer="3651" from-port="2" to-layer="3777" to-port="0" />
+		<edge from-layer="3652" from-port="0" to-layer="3653" to-port="1" />
+		<edge from-layer="3653" from-port="2" to-layer="3655" to-port="0" />
+		<edge from-layer="3654" from-port="0" to-layer="3655" to-port="1" />
+		<edge from-layer="3655" from-port="2" to-layer="3664" to-port="0" />
+		<edge from-layer="3656" from-port="0" to-layer="3657" to-port="0" />
+		<edge from-layer="3657" from-port="1" to-layer="3660" to-port="0" />
+		<edge from-layer="3658" from-port="0" to-layer="3659" to-port="0" />
+		<edge from-layer="3659" from-port="1" to-layer="3660" to-port="1" />
+		<edge from-layer="3660" from-port="2" to-layer="3662" to-port="0" />
+		<edge from-layer="3661" from-port="0" to-layer="3662" to-port="1" />
+		<edge from-layer="3662" from-port="2" to-layer="3663" to-port="0" />
+		<edge from-layer="3663" from-port="1" to-layer="3664" to-port="1" />
+		<edge from-layer="3664" from-port="2" to-layer="3666" to-port="0" />
+		<edge from-layer="3665" from-port="0" to-layer="3666" to-port="1" />
+		<edge from-layer="3666" from-port="2" to-layer="3668" to-port="0" />
+		<edge from-layer="3667" from-port="0" to-layer="3668" to-port="1" />
+		<edge from-layer="3668" from-port="2" to-layer="3672" to-port="0" />
+		<edge from-layer="3668" from-port="2" to-layer="3749" to-port="0" />
+		<edge from-layer="3668" from-port="2" to-layer="3710" to-port="0" />
+		<edge from-layer="3669" from-port="0" to-layer="3672" to-port="1" />
+		<edge from-layer="3670" from-port="0" to-layer="3672" to-port="2" />
+		<edge from-layer="3671" from-port="0" to-layer="3672" to-port="3" />
+		<edge from-layer="3672" from-port="5" to-layer="3674" to-port="0" />
+		<edge from-layer="3673" from-port="0" to-layer="3674" to-port="1" />
+		<edge from-layer="3674" from-port="2" to-layer="3678" to-port="0" />
+		<edge from-layer="3674" from-port="2" to-layer="3696" to-port="0" />
+		<edge from-layer="3675" from-port="0" to-layer="3678" to-port="1" />
+		<edge from-layer="3676" from-port="0" to-layer="3678" to-port="2" />
+		<edge from-layer="3677" from-port="0" to-layer="3678" to-port="3" />
+		<edge from-layer="3678" from-port="5" to-layer="3679" to-port="0" />
+		<edge from-layer="3678" from-port="5" to-layer="3683" to-port="0" />
+		<edge from-layer="3678" from-port="5" to-layer="3689" to-port="0" />
+		<edge from-layer="3679" from-port="2" to-layer="3692" to-port="0" />
+		<edge from-layer="3680" from-port="0" to-layer="3683" to-port="1" />
+		<edge from-layer="3681" from-port="0" to-layer="3683" to-port="2" />
+		<edge from-layer="3682" from-port="0" to-layer="3683" to-port="3" />
+		<edge from-layer="3683" from-port="5" to-layer="3685" to-port="0" />
+		<edge from-layer="3684" from-port="0" to-layer="3685" to-port="1" />
+		<edge from-layer="3685" from-port="2" to-layer="3690" to-port="0" />
+		<edge from-layer="3686" from-port="0" to-layer="3689" to-port="1" />
+		<edge from-layer="3687" from-port="0" to-layer="3689" to-port="2" />
+		<edge from-layer="3688" from-port="0" to-layer="3689" to-port="3" />
+		<edge from-layer="3689" from-port="5" to-layer="3690" to-port="1" />
+		<edge from-layer="3690" from-port="2" to-layer="3691" to-port="0" />
+		<edge from-layer="3691" from-port="2" to-layer="3692" to-port="1" />
+		<edge from-layer="3692" from-port="2" to-layer="3697" to-port="0" />
+		<edge from-layer="3693" from-port="0" to-layer="3696" to-port="1" />
+		<edge from-layer="3694" from-port="0" to-layer="3696" to-port="2" />
+		<edge from-layer="3695" from-port="0" to-layer="3696" to-port="3" />
+		<edge from-layer="3696" from-port="5" to-layer="3697" to-port="1" />
+		<edge from-layer="3697" from-port="2" to-layer="3761" to-port="0" />
+		<edge from-layer="3698" from-port="0" to-layer="3703" to-port="0" />
+		<edge from-layer="3699" from-port="0" to-layer="3702" to-port="1" />
+		<edge from-layer="3700" from-port="0" to-layer="3702" to-port="2" />
+		<edge from-layer="3701" from-port="0" to-layer="3702" to-port="3" />
+		<edge from-layer="3702" from-port="4" to-layer="3703" to-port="1" />
+		<edge from-layer="3703" from-port="2" to-layer="3704" to-port="0" />
+		<edge from-layer="3704" from-port="1" to-layer="3706" to-port="0" />
+		<edge from-layer="3705" from-port="0" to-layer="3706" to-port="2" />
+		<edge from-layer="3706" from-port="3" to-layer="3736" to-port="0" />
+		<edge from-layer="3706" from-port="3" to-layer="3754" to-port="0" />
+		<edge from-layer="3707" from-port="0" to-layer="3710" to-port="1" />
+		<edge from-layer="3708" from-port="0" to-layer="3710" to-port="2" />
+		<edge from-layer="3709" from-port="0" to-layer="3710" to-port="3" />
+		<edge from-layer="3710" from-port="5" to-layer="3712" to-port="0" />
+		<edge from-layer="3711" from-port="0" to-layer="3712" to-port="1" />
+		<edge from-layer="3712" from-port="2" to-layer="3734" to-port="0" />
+		<edge from-layer="3712" from-port="2" to-layer="3716" to-port="0" />
+		<edge from-layer="3713" from-port="0" to-layer="3716" to-port="1" />
+		<edge from-layer="3714" from-port="0" to-layer="3716" to-port="2" />
+		<edge from-layer="3715" from-port="0" to-layer="3716" to-port="3" />
+		<edge from-layer="3716" from-port="5" to-layer="3727" to-port="0" />
+		<edge from-layer="3716" from-port="5" to-layer="3721" to-port="0" />
+		<edge from-layer="3716" from-port="5" to-layer="3717" to-port="0" />
+		<edge from-layer="3717" from-port="2" to-layer="3730" to-port="0" />
+		<edge from-layer="3718" from-port="0" to-layer="3721" to-port="1" />
+		<edge from-layer="3719" from-port="0" to-layer="3721" to-port="2" />
+		<edge from-layer="3720" from-port="0" to-layer="3721" to-port="3" />
+		<edge from-layer="3721" from-port="5" to-layer="3723" to-port="0" />
+		<edge from-layer="3722" from-port="0" to-layer="3723" to-port="1" />
+		<edge from-layer="3723" from-port="2" to-layer="3728" to-port="0" />
+		<edge from-layer="3724" from-port="0" to-layer="3727" to-port="1" />
+		<edge from-layer="3725" from-port="0" to-layer="3727" to-port="2" />
+		<edge from-layer="3726" from-port="0" to-layer="3727" to-port="3" />
+		<edge from-layer="3727" from-port="5" to-layer="3728" to-port="1" />
+		<edge from-layer="3728" from-port="2" to-layer="3729" to-port="0" />
+		<edge from-layer="3729" from-port="2" to-layer="3730" to-port="1" />
+		<edge from-layer="3730" from-port="2" to-layer="3735" to-port="0" />
+		<edge from-layer="3731" from-port="0" to-layer="3734" to-port="1" />
+		<edge from-layer="3732" from-port="0" to-layer="3734" to-port="2" />
+		<edge from-layer="3733" from-port="0" to-layer="3734" to-port="3" />
+		<edge from-layer="3734" from-port="5" to-layer="3735" to-port="1" />
+		<edge from-layer="3735" from-port="2" to-layer="3736" to-port="1" />
+		<edge from-layer="3736" from-port="2" to-layer="5129" to-port="0" />
+		<edge from-layer="3736" from-port="2" to-layer="3761" to-port="1" />
+		<edge from-layer="3737" from-port="0" to-layer="3742" to-port="0" />
+		<edge from-layer="3738" from-port="0" to-layer="3741" to-port="1" />
+		<edge from-layer="3739" from-port="0" to-layer="3741" to-port="2" />
+		<edge from-layer="3740" from-port="0" to-layer="3741" to-port="3" />
+		<edge from-layer="3741" from-port="4" to-layer="3742" to-port="1" />
+		<edge from-layer="3742" from-port="2" to-layer="3743" to-port="0" />
+		<edge from-layer="3743" from-port="1" to-layer="3745" to-port="0" />
+		<edge from-layer="3744" from-port="0" to-layer="3745" to-port="2" />
+		<edge from-layer="3745" from-port="3" to-layer="3752" to-port="0" />
+		<edge from-layer="3746" from-port="0" to-layer="3749" to-port="1" />
+		<edge from-layer="3747" from-port="0" to-layer="3749" to-port="2" />
+		<edge from-layer="3748" from-port="0" to-layer="3749" to-port="3" />
+		<edge from-layer="3749" from-port="5" to-layer="3751" to-port="0" />
+		<edge from-layer="3750" from-port="0" to-layer="3751" to-port="1" />
+		<edge from-layer="3751" from-port="2" to-layer="3752" to-port="1" />
+		<edge from-layer="3752" from-port="2" to-layer="5128" to-port="0" />
+		<edge from-layer="3752" from-port="2" to-layer="3761" to-port="2" />
+		<edge from-layer="3753" from-port="0" to-layer="3760" to-port="1" />
+		<edge from-layer="3754" from-port="1" to-layer="3757" to-port="0" />
+		<edge from-layer="3755" from-port="0" to-layer="3757" to-port="1" />
+		<edge from-layer="3756" from-port="0" to-layer="3757" to-port="2" />
+		<edge from-layer="3757" from-port="3" to-layer="3758" to-port="0" />
+		<edge from-layer="3758" from-port="2" to-layer="3760" to-port="2" />
+		<edge from-layer="3759" from-port="0" to-layer="3760" to-port="3" />
+		<edge from-layer="3760" from-port="5" to-layer="3761" to-port="3" />
+		<edge from-layer="3761" from-port="4" to-layer="3763" to-port="0" />
+		<edge from-layer="3762" from-port="0" to-layer="3763" to-port="1" />
+		<edge from-layer="3763" from-port="2" to-layer="3765" to-port="0" />
+		<edge from-layer="3764" from-port="0" to-layer="3765" to-port="1" />
+		<edge from-layer="3765" from-port="2" to-layer="3774" to-port="0" />
+		<edge from-layer="3766" from-port="0" to-layer="3767" to-port="0" />
+		<edge from-layer="3767" from-port="1" to-layer="3770" to-port="0" />
+		<edge from-layer="3768" from-port="0" to-layer="3769" to-port="0" />
+		<edge from-layer="3769" from-port="1" to-layer="3770" to-port="1" />
+		<edge from-layer="3770" from-port="2" to-layer="3772" to-port="0" />
+		<edge from-layer="3771" from-port="0" to-layer="3772" to-port="1" />
+		<edge from-layer="3772" from-port="2" to-layer="3773" to-port="0" />
+		<edge from-layer="3773" from-port="1" to-layer="3774" to-port="1" />
+		<edge from-layer="3774" from-port="2" to-layer="3776" to-port="0" />
+		<edge from-layer="3775" from-port="0" to-layer="3776" to-port="1" />
+		<edge from-layer="3776" from-port="2" to-layer="3777" to-port="1" />
+		<edge from-layer="3777" from-port="2" to-layer="3778" to-port="0" />
+		<edge from-layer="3778" from-port="2" to-layer="3936" to-port="1" />
+		<edge from-layer="3778" from-port="2" to-layer="3780" to-port="0" />
+		<edge from-layer="3779" from-port="0" to-layer="3780" to-port="1" />
+		<edge from-layer="3780" from-port="2" to-layer="3811" to-port="0" />
+		<edge from-layer="3780" from-port="2" to-layer="3782" to-port="0" />
+		<edge from-layer="3781" from-port="0" to-layer="3782" to-port="1" />
+		<edge from-layer="3782" from-port="2" to-layer="3784" to-port="0" />
+		<edge from-layer="3783" from-port="0" to-layer="3784" to-port="1" />
+		<edge from-layer="3784" from-port="2" to-layer="3793" to-port="0" />
+		<edge from-layer="3785" from-port="0" to-layer="3786" to-port="0" />
+		<edge from-layer="3786" from-port="1" to-layer="3789" to-port="0" />
+		<edge from-layer="3787" from-port="0" to-layer="3788" to-port="0" />
+		<edge from-layer="3788" from-port="1" to-layer="3789" to-port="1" />
+		<edge from-layer="3789" from-port="2" to-layer="3791" to-port="0" />
+		<edge from-layer="3790" from-port="0" to-layer="3791" to-port="1" />
+		<edge from-layer="3791" from-port="2" to-layer="3792" to-port="0" />
+		<edge from-layer="3792" from-port="1" to-layer="3793" to-port="1" />
+		<edge from-layer="3793" from-port="2" to-layer="3796" to-port="0" />
+		<edge from-layer="3794" from-port="0" to-layer="3795" to-port="0" />
+		<edge from-layer="3795" from-port="1" to-layer="3796" to-port="1" />
+		<edge from-layer="3796" from-port="2" to-layer="3797" to-port="0" />
+		<edge from-layer="3797" from-port="1" to-layer="3806" to-port="0" />
+		<edge from-layer="3798" from-port="0" to-layer="3799" to-port="0" />
+		<edge from-layer="3799" from-port="1" to-layer="3802" to-port="0" />
+		<edge from-layer="3800" from-port="0" to-layer="3801" to-port="0" />
+		<edge from-layer="3801" from-port="1" to-layer="3802" to-port="1" />
+		<edge from-layer="3802" from-port="2" to-layer="3804" to-port="0" />
+		<edge from-layer="3803" from-port="0" to-layer="3804" to-port="1" />
+		<edge from-layer="3804" from-port="2" to-layer="3805" to-port="0" />
+		<edge from-layer="3805" from-port="1" to-layer="3806" to-port="1" />
+		<edge from-layer="3806" from-port="2" to-layer="3809" to-port="0" />
+		<edge from-layer="3807" from-port="0" to-layer="3808" to-port="0" />
+		<edge from-layer="3808" from-port="1" to-layer="3809" to-port="1" />
+		<edge from-layer="3809" from-port="2" to-layer="3935" to-port="0" />
+		<edge from-layer="3810" from-port="0" to-layer="3811" to-port="1" />
+		<edge from-layer="3811" from-port="2" to-layer="3813" to-port="0" />
+		<edge from-layer="3812" from-port="0" to-layer="3813" to-port="1" />
+		<edge from-layer="3813" from-port="2" to-layer="3822" to-port="0" />
+		<edge from-layer="3814" from-port="0" to-layer="3815" to-port="0" />
+		<edge from-layer="3815" from-port="1" to-layer="3818" to-port="0" />
+		<edge from-layer="3816" from-port="0" to-layer="3817" to-port="0" />
+		<edge from-layer="3817" from-port="1" to-layer="3818" to-port="1" />
+		<edge from-layer="3818" from-port="2" to-layer="3820" to-port="0" />
+		<edge from-layer="3819" from-port="0" to-layer="3820" to-port="1" />
+		<edge from-layer="3820" from-port="2" to-layer="3821" to-port="0" />
+		<edge from-layer="3821" from-port="1" to-layer="3822" to-port="1" />
+		<edge from-layer="3822" from-port="2" to-layer="3824" to-port="0" />
+		<edge from-layer="3823" from-port="0" to-layer="3824" to-port="1" />
+		<edge from-layer="3824" from-port="2" to-layer="3826" to-port="0" />
+		<edge from-layer="3825" from-port="0" to-layer="3826" to-port="1" />
+		<edge from-layer="3826" from-port="2" to-layer="3830" to-port="0" />
+		<edge from-layer="3826" from-port="2" to-layer="3907" to-port="0" />
+		<edge from-layer="3826" from-port="2" to-layer="3868" to-port="0" />
+		<edge from-layer="3827" from-port="0" to-layer="3830" to-port="1" />
+		<edge from-layer="3828" from-port="0" to-layer="3830" to-port="2" />
+		<edge from-layer="3829" from-port="0" to-layer="3830" to-port="3" />
+		<edge from-layer="3830" from-port="5" to-layer="3832" to-port="0" />
+		<edge from-layer="3831" from-port="0" to-layer="3832" to-port="1" />
+		<edge from-layer="3832" from-port="2" to-layer="3836" to-port="0" />
+		<edge from-layer="3832" from-port="2" to-layer="3854" to-port="0" />
+		<edge from-layer="3833" from-port="0" to-layer="3836" to-port="1" />
+		<edge from-layer="3834" from-port="0" to-layer="3836" to-port="2" />
+		<edge from-layer="3835" from-port="0" to-layer="3836" to-port="3" />
+		<edge from-layer="3836" from-port="5" to-layer="3841" to-port="0" />
+		<edge from-layer="3836" from-port="5" to-layer="3847" to-port="0" />
+		<edge from-layer="3836" from-port="5" to-layer="3837" to-port="0" />
+		<edge from-layer="3837" from-port="2" to-layer="3850" to-port="0" />
+		<edge from-layer="3838" from-port="0" to-layer="3841" to-port="1" />
+		<edge from-layer="3839" from-port="0" to-layer="3841" to-port="2" />
+		<edge from-layer="3840" from-port="0" to-layer="3841" to-port="3" />
+		<edge from-layer="3841" from-port="5" to-layer="3843" to-port="0" />
+		<edge from-layer="3842" from-port="0" to-layer="3843" to-port="1" />
+		<edge from-layer="3843" from-port="2" to-layer="3848" to-port="0" />
+		<edge from-layer="3844" from-port="0" to-layer="3847" to-port="1" />
+		<edge from-layer="3845" from-port="0" to-layer="3847" to-port="2" />
+		<edge from-layer="3846" from-port="0" to-layer="3847" to-port="3" />
+		<edge from-layer="3847" from-port="5" to-layer="3848" to-port="1" />
+		<edge from-layer="3848" from-port="2" to-layer="3849" to-port="0" />
+		<edge from-layer="3849" from-port="2" to-layer="3850" to-port="1" />
+		<edge from-layer="3850" from-port="2" to-layer="3855" to-port="0" />
+		<edge from-layer="3851" from-port="0" to-layer="3854" to-port="1" />
+		<edge from-layer="3852" from-port="0" to-layer="3854" to-port="2" />
+		<edge from-layer="3853" from-port="0" to-layer="3854" to-port="3" />
+		<edge from-layer="3854" from-port="5" to-layer="3855" to-port="1" />
+		<edge from-layer="3855" from-port="2" to-layer="3919" to-port="0" />
+		<edge from-layer="3856" from-port="0" to-layer="3861" to-port="0" />
+		<edge from-layer="3857" from-port="0" to-layer="3860" to-port="1" />
+		<edge from-layer="3858" from-port="0" to-layer="3860" to-port="2" />
+		<edge from-layer="3859" from-port="0" to-layer="3860" to-port="3" />
+		<edge from-layer="3860" from-port="4" to-layer="3861" to-port="1" />
+		<edge from-layer="3861" from-port="2" to-layer="3862" to-port="0" />
+		<edge from-layer="3862" from-port="1" to-layer="3864" to-port="0" />
+		<edge from-layer="3863" from-port="0" to-layer="3864" to-port="2" />
+		<edge from-layer="3864" from-port="3" to-layer="3894" to-port="0" />
+		<edge from-layer="3864" from-port="3" to-layer="3912" to-port="0" />
+		<edge from-layer="3865" from-port="0" to-layer="3868" to-port="1" />
+		<edge from-layer="3866" from-port="0" to-layer="3868" to-port="2" />
+		<edge from-layer="3867" from-port="0" to-layer="3868" to-port="3" />
+		<edge from-layer="3868" from-port="5" to-layer="3870" to-port="0" />
+		<edge from-layer="3869" from-port="0" to-layer="3870" to-port="1" />
+		<edge from-layer="3870" from-port="2" to-layer="3874" to-port="0" />
+		<edge from-layer="3870" from-port="2" to-layer="3892" to-port="0" />
+		<edge from-layer="3871" from-port="0" to-layer="3874" to-port="1" />
+		<edge from-layer="3872" from-port="0" to-layer="3874" to-port="2" />
+		<edge from-layer="3873" from-port="0" to-layer="3874" to-port="3" />
+		<edge from-layer="3874" from-port="5" to-layer="3875" to-port="0" />
+		<edge from-layer="3874" from-port="5" to-layer="3879" to-port="0" />
+		<edge from-layer="3874" from-port="5" to-layer="3885" to-port="0" />
+		<edge from-layer="3875" from-port="2" to-layer="3888" to-port="0" />
+		<edge from-layer="3876" from-port="0" to-layer="3879" to-port="1" />
+		<edge from-layer="3877" from-port="0" to-layer="3879" to-port="2" />
+		<edge from-layer="3878" from-port="0" to-layer="3879" to-port="3" />
+		<edge from-layer="3879" from-port="5" to-layer="3881" to-port="0" />
+		<edge from-layer="3880" from-port="0" to-layer="3881" to-port="1" />
+		<edge from-layer="3881" from-port="2" to-layer="3886" to-port="0" />
+		<edge from-layer="3882" from-port="0" to-layer="3885" to-port="1" />
+		<edge from-layer="3883" from-port="0" to-layer="3885" to-port="2" />
+		<edge from-layer="3884" from-port="0" to-layer="3885" to-port="3" />
+		<edge from-layer="3885" from-port="5" to-layer="3886" to-port="1" />
+		<edge from-layer="3886" from-port="2" to-layer="3887" to-port="0" />
+		<edge from-layer="3887" from-port="2" to-layer="3888" to-port="1" />
+		<edge from-layer="3888" from-port="2" to-layer="3893" to-port="0" />
+		<edge from-layer="3889" from-port="0" to-layer="3892" to-port="1" />
+		<edge from-layer="3890" from-port="0" to-layer="3892" to-port="2" />
 		<edge from-layer="3891" from-port="0" to-layer="3892" to-port="3" />
-		<edge from-layer="3892" from-port="4" to-layer="3897" to-port="1" />
-		<edge from-layer="3893" from-port="0" to-layer="3895" to-port="0" />
-		<edge from-layer="3894" from-port="2" to-layer="3895" to-port="2" />
-		<edge from-layer="3894" from-port="2" to-layer="3902" to-port="2" />
-		<edge from-layer="3895" from-port="4" to-layer="3897" to-port="2" />
-		<edge from-layer="3896" from-port="0" to-layer="3897" to-port="3" />
-		<edge from-layer="3897" from-port="4" to-layer="3904" to-port="0" />
-		<edge from-layer="3898" from-port="0" to-layer="3904" to-port="1" />
-		<edge from-layer="3899" from-port="0" to-layer="3902" to-port="0" />
-		<edge from-layer="3900" from-port="0" to-layer="3902" to-port="1" />
-		<edge from-layer="3901" from-port="0" to-layer="3902" to-port="3" />
-		<edge from-layer="3902" from-port="4" to-layer="3904" to-port="2" />
-		<edge from-layer="3903" from-port="0" to-layer="3904" to-port="3" />
-		<edge from-layer="3904" from-port="4" to-layer="3908" to-port="0" />
-		<edge from-layer="3905" from-port="0" to-layer="3906" to-port="0" />
-		<edge from-layer="3906" from-port="1" to-layer="3908" to-port="1" />
-		<edge from-layer="3907" from-port="0" to-layer="3908" to-port="2" />
-		<edge from-layer="3908" from-port="3" to-layer="3909" to-port="0" />
-		<edge from-layer="3909" from-port="2" to-layer="3917" to-port="0" />
-		<edge from-layer="3909" from-port="2" to-layer="3910" to-port="0" />
-		<edge from-layer="3910" from-port="1" to-layer="3916" to-port="0" />
-		<edge from-layer="3911" from-port="0" to-layer="3913" to-port="1" />
-		<edge from-layer="3912" from-port="0" to-layer="3913" to-port="2" />
-		<edge from-layer="3913" from-port="3" to-layer="3914" to-port="0" />
-		<edge from-layer="3914" from-port="4" to-layer="3915" to-port="0" />
-		<edge from-layer="3915" from-port="1" to-layer="3916" to-port="1" />
-		<edge from-layer="3916" from-port="2" to-layer="3917" to-port="1" />
-		<edge from-layer="3917" from-port="2" to-layer="3918" to-port="3" />
-		<edge from-layer="3918" from-port="4" to-layer="3920" to-port="0" />
-		<edge from-layer="3919" from-port="0" to-layer="3920" to-port="1" />
-		<edge from-layer="3920" from-port="2" to-layer="3922" to-port="0" />
-		<edge from-layer="3921" from-port="0" to-layer="3922" to-port="1" />
-		<edge from-layer="3922" from-port="2" to-layer="3931" to-port="0" />
-		<edge from-layer="3923" from-port="0" to-layer="3924" to-port="0" />
-		<edge from-layer="3924" from-port="1" to-layer="3927" to-port="0" />
-		<edge from-layer="3925" from-port="0" to-layer="3926" to-port="0" />
-		<edge from-layer="3926" from-port="1" to-layer="3927" to-port="1" />
-		<edge from-layer="3927" from-port="2" to-layer="3929" to-port="0" />
-		<edge from-layer="3928" from-port="0" to-layer="3929" to-port="1" />
-		<edge from-layer="3929" from-port="2" to-layer="3930" to-port="0" />
-		<edge from-layer="3930" from-port="1" to-layer="3931" to-port="1" />
-		<edge from-layer="3931" from-port="2" to-layer="3934" to-port="0" />
-		<edge from-layer="3932" from-port="0" to-layer="3933" to-port="0" />
-		<edge from-layer="3933" from-port="1" to-layer="3934" to-port="1" />
+		<edge from-layer="3892" from-port="5" to-layer="3893" to-port="1" />
+		<edge from-layer="3893" from-port="2" to-layer="3894" to-port="1" />
+		<edge from-layer="3894" from-port="2" to-layer="5127" to-port="0" />
+		<edge from-layer="3894" from-port="2" to-layer="3919" to-port="1" />
+		<edge from-layer="3895" from-port="0" to-layer="3900" to-port="0" />
+		<edge from-layer="3896" from-port="0" to-layer="3899" to-port="1" />
+		<edge from-layer="3897" from-port="0" to-layer="3899" to-port="2" />
+		<edge from-layer="3898" from-port="0" to-layer="3899" to-port="3" />
+		<edge from-layer="3899" from-port="4" to-layer="3900" to-port="1" />
+		<edge from-layer="3900" from-port="2" to-layer="3901" to-port="0" />
+		<edge from-layer="3901" from-port="1" to-layer="3903" to-port="0" />
+		<edge from-layer="3902" from-port="0" to-layer="3903" to-port="2" />
+		<edge from-layer="3903" from-port="3" to-layer="3910" to-port="0" />
+		<edge from-layer="3904" from-port="0" to-layer="3907" to-port="1" />
+		<edge from-layer="3905" from-port="0" to-layer="3907" to-port="2" />
+		<edge from-layer="3906" from-port="0" to-layer="3907" to-port="3" />
+		<edge from-layer="3907" from-port="5" to-layer="3909" to-port="0" />
+		<edge from-layer="3908" from-port="0" to-layer="3909" to-port="1" />
+		<edge from-layer="3909" from-port="2" to-layer="3910" to-port="1" />
+		<edge from-layer="3910" from-port="2" to-layer="3919" to-port="2" />
+		<edge from-layer="3910" from-port="2" to-layer="5126" to-port="0" />
+		<edge from-layer="3911" from-port="0" to-layer="3918" to-port="1" />
+		<edge from-layer="3912" from-port="1" to-layer="3915" to-port="0" />
+		<edge from-layer="3913" from-port="0" to-layer="3915" to-port="1" />
+		<edge from-layer="3914" from-port="0" to-layer="3915" to-port="2" />
+		<edge from-layer="3915" from-port="3" to-layer="3916" to-port="0" />
+		<edge from-layer="3916" from-port="2" to-layer="3918" to-port="2" />
+		<edge from-layer="3917" from-port="0" to-layer="3918" to-port="3" />
+		<edge from-layer="3918" from-port="5" to-layer="3919" to-port="3" />
+		<edge from-layer="3919" from-port="4" to-layer="3921" to-port="0" />
+		<edge from-layer="3920" from-port="0" to-layer="3921" to-port="1" />
+		<edge from-layer="3921" from-port="2" to-layer="3923" to-port="0" />
+		<edge from-layer="3922" from-port="0" to-layer="3923" to-port="1" />
+		<edge from-layer="3923" from-port="2" to-layer="3932" to-port="0" />
+		<edge from-layer="3924" from-port="0" to-layer="3925" to-port="0" />
+		<edge from-layer="3925" from-port="1" to-layer="3928" to-port="0" />
+		<edge from-layer="3926" from-port="0" to-layer="3927" to-port="0" />
+		<edge from-layer="3927" from-port="1" to-layer="3928" to-port="1" />
+		<edge from-layer="3928" from-port="2" to-layer="3930" to-port="0" />
+		<edge from-layer="3929" from-port="0" to-layer="3930" to-port="1" />
+		<edge from-layer="3930" from-port="2" to-layer="3931" to-port="0" />
+		<edge from-layer="3931" from-port="1" to-layer="3932" to-port="1" />
+		<edge from-layer="3932" from-port="2" to-layer="3934" to-port="0" />
+		<edge from-layer="3933" from-port="0" to-layer="3934" to-port="1" />
 		<edge from-layer="3934" from-port="2" to-layer="3935" to-port="1" />
 		<edge from-layer="3935" from-port="2" to-layer="3936" to-port="0" />
+		<edge from-layer="3936" from-port="2" to-layer="4094" to-port="1" />
 		<edge from-layer="3936" from-port="2" to-layer="3938" to-port="0" />
-		<edge from-layer="3936" from-port="2" to-layer="4178" to-port="1" />
 		<edge from-layer="3937" from-port="0" to-layer="3938" to-port="1" />
-		<edge from-layer="3938" from-port="2" to-layer="3941" to-port="0" />
-		<edge from-layer="3938" from-port="2" to-layer="3972" to-port="0" />
-		<edge from-layer="3939" from-port="0" to-layer="3940" to-port="0" />
-		<edge from-layer="3940" from-port="1" to-layer="3941" to-port="1" />
-		<edge from-layer="3941" from-port="2" to-layer="3944" to-port="0" />
-		<edge from-layer="3942" from-port="0" to-layer="3943" to-port="0" />
-		<edge from-layer="3943" from-port="1" to-layer="3944" to-port="1" />
-		<edge from-layer="3944" from-port="2" to-layer="3953" to-port="0" />
+		<edge from-layer="3938" from-port="2" to-layer="3940" to-port="0" />
+		<edge from-layer="3938" from-port="2" to-layer="3969" to-port="0" />
+		<edge from-layer="3939" from-port="0" to-layer="3940" to-port="1" />
+		<edge from-layer="3940" from-port="2" to-layer="3942" to-port="0" />
+		<edge from-layer="3941" from-port="0" to-layer="3942" to-port="1" />
+		<edge from-layer="3942" from-port="2" to-layer="3951" to-port="0" />
+		<edge from-layer="3943" from-port="0" to-layer="3944" to-port="0" />
+		<edge from-layer="3944" from-port="1" to-layer="3947" to-port="0" />
 		<edge from-layer="3945" from-port="0" to-layer="3946" to-port="0" />
-		<edge from-layer="3946" from-port="1" to-layer="3949" to-port="0" />
-		<edge from-layer="3947" from-port="0" to-layer="3948" to-port="0" />
-		<edge from-layer="3948" from-port="1" to-layer="3949" to-port="1" />
-		<edge from-layer="3949" from-port="2" to-layer="3951" to-port="0" />
-		<edge from-layer="3950" from-port="0" to-layer="3951" to-port="1" />
-		<edge from-layer="3951" from-port="2" to-layer="3952" to-port="0" />
-		<edge from-layer="3952" from-port="1" to-layer="3953" to-port="1" />
-		<edge from-layer="3953" from-port="2" to-layer="3956" to-port="0" />
-		<edge from-layer="3954" from-port="0" to-layer="3955" to-port="0" />
-		<edge from-layer="3955" from-port="1" to-layer="3956" to-port="1" />
-		<edge from-layer="3956" from-port="2" to-layer="3957" to-port="0" />
-		<edge from-layer="3957" from-port="1" to-layer="3966" to-port="0" />
+		<edge from-layer="3946" from-port="1" to-layer="3947" to-port="1" />
+		<edge from-layer="3947" from-port="2" to-layer="3949" to-port="0" />
+		<edge from-layer="3948" from-port="0" to-layer="3949" to-port="1" />
+		<edge from-layer="3949" from-port="2" to-layer="3950" to-port="0" />
+		<edge from-layer="3950" from-port="1" to-layer="3951" to-port="1" />
+		<edge from-layer="3951" from-port="2" to-layer="3954" to-port="0" />
+		<edge from-layer="3952" from-port="0" to-layer="3953" to-port="0" />
+		<edge from-layer="3953" from-port="1" to-layer="3954" to-port="1" />
+		<edge from-layer="3954" from-port="2" to-layer="3955" to-port="0" />
+		<edge from-layer="3955" from-port="1" to-layer="3964" to-port="0" />
+		<edge from-layer="3956" from-port="0" to-layer="3957" to-port="0" />
+		<edge from-layer="3957" from-port="1" to-layer="3960" to-port="0" />
 		<edge from-layer="3958" from-port="0" to-layer="3959" to-port="0" />
-		<edge from-layer="3959" from-port="1" to-layer="3962" to-port="0" />
-		<edge from-layer="3960" from-port="0" to-layer="3961" to-port="0" />
-		<edge from-layer="3961" from-port="1" to-layer="3962" to-port="1" />
-		<edge from-layer="3962" from-port="2" to-layer="3964" to-port="0" />
-		<edge from-layer="3963" from-port="0" to-layer="3964" to-port="1" />
-		<edge from-layer="3964" from-port="2" to-layer="3965" to-port="0" />
-		<edge from-layer="3965" from-port="1" to-layer="3966" to-port="1" />
-		<edge from-layer="3966" from-port="2" to-layer="3969" to-port="0" />
-		<edge from-layer="3967" from-port="0" to-layer="3968" to-port="0" />
-		<edge from-layer="3968" from-port="1" to-layer="3969" to-port="1" />
-		<edge from-layer="3969" from-port="2" to-layer="4177" to-port="0" />
-		<edge from-layer="3970" from-port="0" to-layer="3971" to-port="0" />
-		<edge from-layer="3971" from-port="1" to-layer="3972" to-port="1" />
-		<edge from-layer="3972" from-port="2" to-layer="3975" to-port="0" />
-		<edge from-layer="3973" from-port="0" to-layer="3974" to-port="0" />
-		<edge from-layer="3974" from-port="1" to-layer="3975" to-port="1" />
-		<edge from-layer="3975" from-port="2" to-layer="3984" to-port="0" />
-		<edge from-layer="3976" from-port="0" to-layer="3977" to-port="0" />
-		<edge from-layer="3977" from-port="1" to-layer="3980" to-port="0" />
-		<edge from-layer="3978" from-port="0" to-layer="3979" to-port="0" />
+		<edge from-layer="3959" from-port="1" to-layer="3960" to-port="1" />
+		<edge from-layer="3960" from-port="2" to-layer="3962" to-port="0" />
+		<edge from-layer="3961" from-port="0" to-layer="3962" to-port="1" />
+		<edge from-layer="3962" from-port="2" to-layer="3963" to-port="0" />
+		<edge from-layer="3963" from-port="1" to-layer="3964" to-port="1" />
+		<edge from-layer="3964" from-port="2" to-layer="3967" to-port="0" />
+		<edge from-layer="3965" from-port="0" to-layer="3966" to-port="0" />
+		<edge from-layer="3966" from-port="1" to-layer="3967" to-port="1" />
+		<edge from-layer="3967" from-port="2" to-layer="4093" to-port="0" />
+		<edge from-layer="3968" from-port="0" to-layer="3969" to-port="1" />
+		<edge from-layer="3969" from-port="2" to-layer="3971" to-port="0" />
+		<edge from-layer="3970" from-port="0" to-layer="3971" to-port="1" />
+		<edge from-layer="3971" from-port="2" to-layer="3980" to-port="0" />
+		<edge from-layer="3972" from-port="0" to-layer="3973" to-port="0" />
+		<edge from-layer="3973" from-port="1" to-layer="3976" to-port="0" />
+		<edge from-layer="3974" from-port="0" to-layer="3975" to-port="0" />
+		<edge from-layer="3975" from-port="1" to-layer="3976" to-port="1" />
+		<edge from-layer="3976" from-port="2" to-layer="3978" to-port="0" />
+		<edge from-layer="3977" from-port="0" to-layer="3978" to-port="1" />
+		<edge from-layer="3978" from-port="2" to-layer="3979" to-port="0" />
 		<edge from-layer="3979" from-port="1" to-layer="3980" to-port="1" />
 		<edge from-layer="3980" from-port="2" to-layer="3982" to-port="0" />
 		<edge from-layer="3981" from-port="0" to-layer="3982" to-port="1" />
-		<edge from-layer="3982" from-port="2" to-layer="3983" to-port="0" />
-		<edge from-layer="3983" from-port="1" to-layer="3984" to-port="1" />
-		<edge from-layer="3984" from-port="2" to-layer="3987" to-port="0" />
-		<edge from-layer="3985" from-port="0" to-layer="3986" to-port="0" />
-		<edge from-layer="3986" from-port="1" to-layer="3987" to-port="1" />
-		<edge from-layer="3987" from-port="2" to-layer="3989" to-port="0" />
-		<edge from-layer="3988" from-port="0" to-layer="3989" to-port="1" />
-		<edge from-layer="3989" from-port="2" to-layer="4006" to-port="0" />
-		<edge from-layer="3989" from-port="2" to-layer="3993" to-port="0" />
-		<edge from-layer="3989" from-port="2" to-layer="4117" to-port="0" />
-		<edge from-layer="3990" from-port="0" to-layer="3993" to-port="1" />
-		<edge from-layer="3991" from-port="0" to-layer="3993" to-port="2" />
-		<edge from-layer="3992" from-port="0" to-layer="3993" to-port="3" />
-		<edge from-layer="3993" from-port="4" to-layer="3995" to-port="0" />
-		<edge from-layer="3994" from-port="0" to-layer="3995" to-port="1" />
-		<edge from-layer="3995" from-port="2" to-layer="3999" to-port="0" />
-		<edge from-layer="3995" from-port="2" to-layer="4068" to-port="0" />
+		<edge from-layer="3982" from-port="2" to-layer="3984" to-port="0" />
+		<edge from-layer="3983" from-port="0" to-layer="3984" to-port="1" />
+		<edge from-layer="3984" from-port="2" to-layer="4065" to-port="0" />
+		<edge from-layer="3984" from-port="2" to-layer="4026" to-port="0" />
+		<edge from-layer="3984" from-port="2" to-layer="3988" to-port="0" />
+		<edge from-layer="3985" from-port="0" to-layer="3988" to-port="1" />
+		<edge from-layer="3986" from-port="0" to-layer="3988" to-port="2" />
+		<edge from-layer="3987" from-port="0" to-layer="3988" to-port="3" />
+		<edge from-layer="3988" from-port="5" to-layer="3990" to-port="0" />
+		<edge from-layer="3989" from-port="0" to-layer="3990" to-port="1" />
+		<edge from-layer="3990" from-port="2" to-layer="4012" to-port="0" />
+		<edge from-layer="3990" from-port="2" to-layer="3994" to-port="0" />
+		<edge from-layer="3991" from-port="0" to-layer="3994" to-port="1" />
+		<edge from-layer="3992" from-port="0" to-layer="3994" to-port="2" />
+		<edge from-layer="3993" from-port="0" to-layer="3994" to-port="3" />
+		<edge from-layer="3994" from-port="5" to-layer="4005" to-port="0" />
+		<edge from-layer="3994" from-port="5" to-layer="3999" to-port="0" />
+		<edge from-layer="3994" from-port="5" to-layer="3995" to-port="0" />
+		<edge from-layer="3995" from-port="2" to-layer="4008" to-port="0" />
 		<edge from-layer="3996" from-port="0" to-layer="3999" to-port="1" />
 		<edge from-layer="3997" from-port="0" to-layer="3999" to-port="2" />
 		<edge from-layer="3998" from-port="0" to-layer="3999" to-port="3" />
-		<edge from-layer="3999" from-port="4" to-layer="4033" to-port="0" />
-		<edge from-layer="3999" from-port="4" to-layer="4053" to-port="0" />
-		<edge from-layer="3999" from-port="4" to-layer="4047" to-port="0" />
-		<edge from-layer="3999" from-port="4" to-layer="4036" to-port="0" />
-		<edge from-layer="4000" from-port="0" to-layer="4001" to-port="0" />
-		<edge from-layer="4001" from-port="1" to-layer="4029" to-port="0" />
-		<edge from-layer="4002" from-port="0" to-layer="4029" to-port="1" />
-		<edge from-layer="4003" from-port="0" to-layer="4006" to-port="1" />
-		<edge from-layer="4004" from-port="0" to-layer="4006" to-port="2" />
-		<edge from-layer="4005" from-port="0" to-layer="4006" to-port="3" />
-		<edge from-layer="4006" from-port="4" to-layer="4008" to-port="0" />
-		<edge from-layer="4007" from-port="0" to-layer="4008" to-port="1" />
-		<edge from-layer="4008" from-port="2" to-layer="4009" to-port="0" />
-		<edge from-layer="4008" from-port="2" to-layer="4073" to-port="0" />
-		<edge from-layer="4008" from-port="2" to-layer="4101" to-port="0" />
-		<edge from-layer="4009" from-port="1" to-layer="4011" to-port="0" />
-		<edge from-layer="4010" from-port="0" to-layer="4011" to-port="2" />
-		<edge from-layer="4011" from-port="3" to-layer="4025" to-port="0" />
-		<edge from-layer="4012" from-port="0" to-layer="4013" to-port="0" />
-		<edge from-layer="4013" from-port="1" to-layer="4018" to-port="0" />
-		<edge from-layer="4014" from-port="0" to-layer="4017" to-port="1" />
-		<edge from-layer="4015" from-port="0" to-layer="4017" to-port="2" />
-		<edge from-layer="4016" from-port="0" to-layer="4017" to-port="3" />
-		<edge from-layer="4017" from-port="4" to-layer="4018" to-port="1" />
-		<edge from-layer="4018" from-port="2" to-layer="4019" to-port="0" />
-		<edge from-layer="4019" from-port="1" to-layer="4021" to-port="0" />
-		<edge from-layer="4020" from-port="0" to-layer="4021" to-port="2" />
-		<edge from-layer="4021" from-port="3" to-layer="4103" to-port="0" />
-		<edge from-layer="4021" from-port="3" to-layer="4022" to-port="0" />
-		<edge from-layer="4022" from-port="1" to-layer="4024" to-port="0" />
-		<edge from-layer="4023" from-port="0" to-layer="4024" to-port="2" />
-		<edge from-layer="4024" from-port="3" to-layer="4025" to-port="1" />
-		<edge from-layer="4025" from-port="2" to-layer="4027" to-port="0" />
-		<edge from-layer="4026" from-port="0" to-layer="4027" to-port="1" />
-		<edge from-layer="4027" from-port="2" to-layer="4029" to-port="2" />
-		<edge from-layer="4027" from-port="2" to-layer="4059" to-port="2" />
-		<edge from-layer="4028" from-port="0" to-layer="4029" to-port="3" />
-		<edge from-layer="4029" from-port="4" to-layer="4031" to-port="0" />
-		<edge from-layer="4030" from-port="0" to-layer="4031" to-port="2" />
-		<edge from-layer="4031" from-port="3" to-layer="4032" to-port="0" />
-		<edge from-layer="4032" from-port="2" to-layer="4033" to-port="1" />
-		<edge from-layer="4032" from-port="2" to-layer="4074" to-port="1" />
-		<edge from-layer="4033" from-port="2" to-layer="4064" to-port="0" />
-		<edge from-layer="4034" from-port="0" to-layer="4044" to-port="0" />
-		<edge from-layer="4035" from-port="0" to-layer="4044" to-port="1" />
-		<edge from-layer="4036" from-port="1" to-layer="4038" to-port="0" />
-		<edge from-layer="4037" from-port="0" to-layer="4038" to-port="2" />
-		<edge from-layer="4038" from-port="3" to-layer="4039" to-port="0" />
-		<edge from-layer="4039" from-port="2" to-layer="4040" to-port="0" />
-		<edge from-layer="4040" from-port="1" to-layer="4042" to-port="0" />
-		<edge from-layer="4041" from-port="0" to-layer="4042" to-port="1" />
-		<edge from-layer="4042" from-port="2" to-layer="4044" to-port="2" />
-		<edge from-layer="4043" from-port="0" to-layer="4044" to-port="3" />
-		<edge from-layer="4044" from-port="4" to-layer="4053" to-port="2" />
-		<edge from-layer="4044" from-port="4" to-layer="4047" to-port="1" />
-		<edge from-layer="4045" from-port="0" to-layer="4047" to-port="2" />
-		<edge from-layer="4046" from-port="0" to-layer="4047" to-port="3" />
-		<edge from-layer="4047" from-port="4" to-layer="4050" to-port="0" />
-		<edge from-layer="4048" from-port="0" to-layer="4049" to-port="0" />
-		<edge from-layer="4049" from-port="1" to-layer="4050" to-port="1" />
-		<edge from-layer="4050" from-port="2" to-layer="4054" to-port="0" />
-		<edge from-layer="4051" from-port="0" to-layer="4053" to-port="1" />
-		<edge from-layer="4052" from-port="0" to-layer="4053" to-port="3" />
-		<edge from-layer="4053" from-port="4" to-layer="4054" to-port="1" />
-		<edge from-layer="4054" from-port="2" to-layer="4063" to-port="0" />
-		<edge from-layer="4055" from-port="0" to-layer="4056" to-port="0" />
-		<edge from-layer="4056" from-port="1" to-layer="4059" to-port="0" />
-		<edge from-layer="4057" from-port="0" to-layer="4059" to-port="1" />
-		<edge from-layer="4058" from-port="0" to-layer="4059" to-port="3" />
-		<edge from-layer="4059" from-port="4" to-layer="4061" to-port="0" />
+		<edge from-layer="3999" from-port="5" to-layer="4001" to-port="0" />
+		<edge from-layer="4000" from-port="0" to-layer="4001" to-port="1" />
+		<edge from-layer="4001" from-port="2" to-layer="4006" to-port="0" />
+		<edge from-layer="4002" from-port="0" to-layer="4005" to-port="1" />
+		<edge from-layer="4003" from-port="0" to-layer="4005" to-port="2" />
+		<edge from-layer="4004" from-port="0" to-layer="4005" to-port="3" />
+		<edge from-layer="4005" from-port="5" to-layer="4006" to-port="1" />
+		<edge from-layer="4006" from-port="2" to-layer="4007" to-port="0" />
+		<edge from-layer="4007" from-port="2" to-layer="4008" to-port="1" />
+		<edge from-layer="4008" from-port="2" to-layer="4013" to-port="0" />
+		<edge from-layer="4009" from-port="0" to-layer="4012" to-port="1" />
+		<edge from-layer="4010" from-port="0" to-layer="4012" to-port="2" />
+		<edge from-layer="4011" from-port="0" to-layer="4012" to-port="3" />
+		<edge from-layer="4012" from-port="5" to-layer="4013" to-port="1" />
+		<edge from-layer="4013" from-port="2" to-layer="4077" to-port="0" />
+		<edge from-layer="4014" from-port="0" to-layer="4019" to-port="0" />
+		<edge from-layer="4015" from-port="0" to-layer="4018" to-port="1" />
+		<edge from-layer="4016" from-port="0" to-layer="4018" to-port="2" />
+		<edge from-layer="4017" from-port="0" to-layer="4018" to-port="3" />
+		<edge from-layer="4018" from-port="4" to-layer="4019" to-port="1" />
+		<edge from-layer="4019" from-port="2" to-layer="4020" to-port="0" />
+		<edge from-layer="4020" from-port="1" to-layer="4022" to-port="0" />
+		<edge from-layer="4021" from-port="0" to-layer="4022" to-port="2" />
+		<edge from-layer="4022" from-port="3" to-layer="4070" to-port="0" />
+		<edge from-layer="4022" from-port="3" to-layer="4052" to-port="0" />
+		<edge from-layer="4023" from-port="0" to-layer="4026" to-port="1" />
+		<edge from-layer="4024" from-port="0" to-layer="4026" to-port="2" />
+		<edge from-layer="4025" from-port="0" to-layer="4026" to-port="3" />
+		<edge from-layer="4026" from-port="5" to-layer="4028" to-port="0" />
+		<edge from-layer="4027" from-port="0" to-layer="4028" to-port="1" />
+		<edge from-layer="4028" from-port="2" to-layer="4032" to-port="0" />
+		<edge from-layer="4028" from-port="2" to-layer="4050" to-port="0" />
+		<edge from-layer="4029" from-port="0" to-layer="4032" to-port="1" />
+		<edge from-layer="4030" from-port="0" to-layer="4032" to-port="2" />
+		<edge from-layer="4031" from-port="0" to-layer="4032" to-port="3" />
+		<edge from-layer="4032" from-port="5" to-layer="4033" to-port="0" />
+		<edge from-layer="4032" from-port="5" to-layer="4037" to-port="0" />
+		<edge from-layer="4032" from-port="5" to-layer="4043" to-port="0" />
+		<edge from-layer="4033" from-port="2" to-layer="4046" to-port="0" />
+		<edge from-layer="4034" from-port="0" to-layer="4037" to-port="1" />
+		<edge from-layer="4035" from-port="0" to-layer="4037" to-port="2" />
+		<edge from-layer="4036" from-port="0" to-layer="4037" to-port="3" />
+		<edge from-layer="4037" from-port="5" to-layer="4039" to-port="0" />
+		<edge from-layer="4038" from-port="0" to-layer="4039" to-port="1" />
+		<edge from-layer="4039" from-port="2" to-layer="4044" to-port="0" />
+		<edge from-layer="4040" from-port="0" to-layer="4043" to-port="1" />
+		<edge from-layer="4041" from-port="0" to-layer="4043" to-port="2" />
+		<edge from-layer="4042" from-port="0" to-layer="4043" to-port="3" />
+		<edge from-layer="4043" from-port="5" to-layer="4044" to-port="1" />
+		<edge from-layer="4044" from-port="2" to-layer="4045" to-port="0" />
+		<edge from-layer="4045" from-port="2" to-layer="4046" to-port="1" />
+		<edge from-layer="4046" from-port="2" to-layer="4051" to-port="0" />
+		<edge from-layer="4047" from-port="0" to-layer="4050" to-port="1" />
+		<edge from-layer="4048" from-port="0" to-layer="4050" to-port="2" />
+		<edge from-layer="4049" from-port="0" to-layer="4050" to-port="3" />
+		<edge from-layer="4050" from-port="5" to-layer="4051" to-port="1" />
+		<edge from-layer="4051" from-port="2" to-layer="4052" to-port="1" />
+		<edge from-layer="4052" from-port="2" to-layer="4077" to-port="1" />
+		<edge from-layer="4052" from-port="2" to-layer="5125" to-port="0" />
+		<edge from-layer="4053" from-port="0" to-layer="4058" to-port="0" />
+		<edge from-layer="4054" from-port="0" to-layer="4057" to-port="1" />
+		<edge from-layer="4055" from-port="0" to-layer="4057" to-port="2" />
+		<edge from-layer="4056" from-port="0" to-layer="4057" to-port="3" />
+		<edge from-layer="4057" from-port="4" to-layer="4058" to-port="1" />
+		<edge from-layer="4058" from-port="2" to-layer="4059" to-port="0" />
+		<edge from-layer="4059" from-port="1" to-layer="4061" to-port="0" />
 		<edge from-layer="4060" from-port="0" to-layer="4061" to-port="2" />
-		<edge from-layer="4061" from-port="3" to-layer="4062" to-port="0" />
-		<edge from-layer="4062" from-port="2" to-layer="4063" to-port="1" />
-		<edge from-layer="4062" from-port="2" to-layer="4096" to-port="1" />
-		<edge from-layer="4063" from-port="2" to-layer="4064" to-port="1" />
-		<edge from-layer="4064" from-port="2" to-layer="4069" to-port="0" />
-		<edge from-layer="4065" from-port="0" to-layer="4068" to-port="1" />
-		<edge from-layer="4066" from-port="0" to-layer="4068" to-port="2" />
-		<edge from-layer="4067" from-port="0" to-layer="4068" to-port="3" />
-		<edge from-layer="4068" from-port="4" to-layer="4069" to-port="1" />
-		<edge from-layer="4069" from-port="2" to-layer="4127" to-port="0" />
-		<edge from-layer="4069" from-port="2" to-layer="4160" to-port="0" />
-		<edge from-layer="4070" from-port="0" to-layer="4073" to-port="1" />
-		<edge from-layer="4071" from-port="0" to-layer="4073" to-port="2" />
-		<edge from-layer="4072" from-port="0" to-layer="4073" to-port="3" />
-		<edge from-layer="4073" from-port="4" to-layer="4094" to-port="0" />
-		<edge from-layer="4073" from-port="4" to-layer="4077" to-port="0" />
-		<edge from-layer="4073" from-port="4" to-layer="4074" to-port="0" />
-		<edge from-layer="4073" from-port="4" to-layer="4088" to-port="0" />
-		<edge from-layer="4074" from-port="2" to-layer="4097" to-port="0" />
-		<edge from-layer="4075" from-port="0" to-layer="4085" to-port="0" />
-		<edge from-layer="4076" from-port="0" to-layer="4085" to-port="1" />
-		<edge from-layer="4077" from-port="1" to-layer="4079" to-port="0" />
-		<edge from-layer="4078" from-port="0" to-layer="4079" to-port="2" />
-		<edge from-layer="4079" from-port="3" to-layer="4080" to-port="0" />
-		<edge from-layer="4080" from-port="2" to-layer="4081" to-port="0" />
-		<edge from-layer="4081" from-port="1" to-layer="4083" to-port="0" />
-		<edge from-layer="4082" from-port="0" to-layer="4083" to-port="1" />
-		<edge from-layer="4083" from-port="2" to-layer="4085" to-port="2" />
-		<edge from-layer="4084" from-port="0" to-layer="4085" to-port="3" />
-		<edge from-layer="4085" from-port="4" to-layer="4088" to-port="1" />
-		<edge from-layer="4085" from-port="4" to-layer="4094" to-port="2" />
-		<edge from-layer="4086" from-port="0" to-layer="4088" to-port="2" />
-		<edge from-layer="4087" from-port="0" to-layer="4088" to-port="3" />
-		<edge from-layer="4088" from-port="4" to-layer="4091" to-port="0" />
-		<edge from-layer="4089" from-port="0" to-layer="4090" to-port="0" />
-		<edge from-layer="4090" from-port="1" to-layer="4091" to-port="1" />
-		<edge from-layer="4091" from-port="2" to-layer="4095" to-port="0" />
-		<edge from-layer="4092" from-port="0" to-layer="4094" to-port="1" />
-		<edge from-layer="4093" from-port="0" to-layer="4094" to-port="3" />
-		<edge from-layer="4094" from-port="4" to-layer="4095" to-port="1" />
-		<edge from-layer="4095" from-port="2" to-layer="4096" to-port="0" />
-		<edge from-layer="4096" from-port="2" to-layer="4097" to-port="1" />
-		<edge from-layer="4097" from-port="2" to-layer="4102" to-port="0" />
-		<edge from-layer="4098" from-port="0" to-layer="4101" to-port="1" />
-		<edge from-layer="4099" from-port="0" to-layer="4101" to-port="2" />
-		<edge from-layer="4100" from-port="0" to-layer="4101" to-port="3" />
-		<edge from-layer="4101" from-port="4" to-layer="4102" to-port="1" />
-		<edge from-layer="4102" from-port="2" to-layer="4103" to-port="1" />
-		<edge from-layer="4103" from-port="2" to-layer="7706" to-port="0" />
-		<edge from-layer="4103" from-port="2" to-layer="4124" to-port="0" />
-		<edge from-layer="4103" from-port="2" to-layer="4160" to-port="1" />
-		<edge from-layer="4104" from-port="0" to-layer="4105" to-port="0" />
-		<edge from-layer="4105" from-port="1" to-layer="4110" to-port="0" />
-		<edge from-layer="4106" from-port="0" to-layer="4109" to-port="1" />
-		<edge from-layer="4107" from-port="0" to-layer="4109" to-port="2" />
-		<edge from-layer="4108" from-port="0" to-layer="4109" to-port="3" />
-		<edge from-layer="4109" from-port="4" to-layer="4110" to-port="1" />
-		<edge from-layer="4110" from-port="2" to-layer="4111" to-port="0" />
-		<edge from-layer="4111" from-port="1" to-layer="4113" to-port="0" />
-		<edge from-layer="4112" from-port="0" to-layer="4113" to-port="2" />
-		<edge from-layer="4113" from-port="3" to-layer="4120" to-port="0" />
-		<edge from-layer="4114" from-port="0" to-layer="4117" to-port="1" />
-		<edge from-layer="4115" from-port="0" to-layer="4117" to-port="2" />
-		<edge from-layer="4116" from-port="0" to-layer="4117" to-port="3" />
-		<edge from-layer="4117" from-port="4" to-layer="4119" to-port="0" />
-		<edge from-layer="4118" from-port="0" to-layer="4119" to-port="1" />
-		<edge from-layer="4119" from-port="2" to-layer="4120" to-port="1" />
-		<edge from-layer="4120" from-port="2" to-layer="7705" to-port="0" />
-		<edge from-layer="4120" from-port="2" to-layer="4160" to-port="2" />
-		<edge from-layer="4121" from-port="0" to-layer="4139" to-port="0" />
-		<edge from-layer="4122" from-port="0" to-layer="4134" to-port="0" />
-		<edge from-layer="4123" from-port="0" to-layer="4134" to-port="1" />
-		<edge from-layer="4123" from-port="0" to-layer="4137" to-port="1" />
-		<edge from-layer="4124" from-port="1" to-layer="4126" to-port="0" />
-		<edge from-layer="4125" from-port="0" to-layer="4126" to-port="2" />
-		<edge from-layer="4126" from-port="3" to-layer="4130" to-port="0" />
-		<edge from-layer="4126" from-port="3" to-layer="4136" to-port="0" />
-		<edge from-layer="4127" from-port="1" to-layer="4155" to-port="0" />
-		<edge from-layer="4127" from-port="1" to-layer="4129" to-port="0" />
-		<edge from-layer="4128" from-port="0" to-layer="4129" to-port="2" />
-		<edge from-layer="4129" from-port="3" to-layer="4130" to-port="1" />
-		<edge from-layer="4130" from-port="2" to-layer="4132" to-port="0" />
-		<edge from-layer="4131" from-port="0" to-layer="4136" to-port="1" />
-		<edge from-layer="4131" from-port="0" to-layer="4132" to-port="1" />
-		<edge from-layer="4132" from-port="2" to-layer="4134" to-port="2" />
-		<edge from-layer="4133" from-port="0" to-layer="4134" to-port="3" />
-		<edge from-layer="4133" from-port="0" to-layer="4137" to-port="3" />
-		<edge from-layer="4134" from-port="4" to-layer="4139" to-port="1" />
-		<edge from-layer="4135" from-port="0" to-layer="4137" to-port="0" />
-		<edge from-layer="4136" from-port="2" to-layer="4137" to-port="2" />
-		<edge from-layer="4136" from-port="2" to-layer="4144" to-port="2" />
-		<edge from-layer="4137" from-port="4" to-layer="4139" to-port="2" />
-		<edge from-layer="4138" from-port="0" to-layer="4139" to-port="3" />
-		<edge from-layer="4139" from-port="4" to-layer="4146" to-port="0" />
-		<edge from-layer="4140" from-port="0" to-layer="4146" to-port="1" />
-		<edge from-layer="4141" from-port="0" to-layer="4144" to-port="0" />
-		<edge from-layer="4142" from-port="0" to-layer="4144" to-port="1" />
-		<edge from-layer="4143" from-port="0" to-layer="4144" to-port="3" />
-		<edge from-layer="4144" from-port="4" to-layer="4146" to-port="2" />
+		<edge from-layer="4061" from-port="3" to-layer="4068" to-port="0" />
+		<edge from-layer="4062" from-port="0" to-layer="4065" to-port="1" />
+		<edge from-layer="4063" from-port="0" to-layer="4065" to-port="2" />
+		<edge from-layer="4064" from-port="0" to-layer="4065" to-port="3" />
+		<edge from-layer="4065" from-port="5" to-layer="4067" to-port="0" />
+		<edge from-layer="4066" from-port="0" to-layer="4067" to-port="1" />
+		<edge from-layer="4067" from-port="2" to-layer="4068" to-port="1" />
+		<edge from-layer="4068" from-port="2" to-layer="4077" to-port="2" />
+		<edge from-layer="4068" from-port="2" to-layer="5124" to-port="0" />
+		<edge from-layer="4069" from-port="0" to-layer="4076" to-port="1" />
+		<edge from-layer="4070" from-port="1" to-layer="4073" to-port="0" />
+		<edge from-layer="4071" from-port="0" to-layer="4073" to-port="1" />
+		<edge from-layer="4072" from-port="0" to-layer="4073" to-port="2" />
+		<edge from-layer="4073" from-port="3" to-layer="4074" to-port="0" />
+		<edge from-layer="4074" from-port="2" to-layer="4076" to-port="2" />
+		<edge from-layer="4075" from-port="0" to-layer="4076" to-port="3" />
+		<edge from-layer="4076" from-port="5" to-layer="4077" to-port="3" />
+		<edge from-layer="4077" from-port="4" to-layer="4079" to-port="0" />
+		<edge from-layer="4078" from-port="0" to-layer="4079" to-port="1" />
+		<edge from-layer="4079" from-port="2" to-layer="4081" to-port="0" />
+		<edge from-layer="4080" from-port="0" to-layer="4081" to-port="1" />
+		<edge from-layer="4081" from-port="2" to-layer="4090" to-port="0" />
+		<edge from-layer="4082" from-port="0" to-layer="4083" to-port="0" />
+		<edge from-layer="4083" from-port="1" to-layer="4086" to-port="0" />
+		<edge from-layer="4084" from-port="0" to-layer="4085" to-port="0" />
+		<edge from-layer="4085" from-port="1" to-layer="4086" to-port="1" />
+		<edge from-layer="4086" from-port="2" to-layer="4088" to-port="0" />
+		<edge from-layer="4087" from-port="0" to-layer="4088" to-port="1" />
+		<edge from-layer="4088" from-port="2" to-layer="4089" to-port="0" />
+		<edge from-layer="4089" from-port="1" to-layer="4090" to-port="1" />
+		<edge from-layer="4090" from-port="2" to-layer="4092" to-port="0" />
+		<edge from-layer="4091" from-port="0" to-layer="4092" to-port="1" />
+		<edge from-layer="4092" from-port="2" to-layer="4093" to-port="1" />
+		<edge from-layer="4093" from-port="2" to-layer="4094" to-port="0" />
+		<edge from-layer="4094" from-port="2" to-layer="4096" to-port="0" />
+		<edge from-layer="4094" from-port="2" to-layer="4252" to-port="1" />
+		<edge from-layer="4095" from-port="0" to-layer="4096" to-port="1" />
+		<edge from-layer="4096" from-port="2" to-layer="4127" to-port="0" />
+		<edge from-layer="4096" from-port="2" to-layer="4098" to-port="0" />
+		<edge from-layer="4097" from-port="0" to-layer="4098" to-port="1" />
+		<edge from-layer="4098" from-port="2" to-layer="4100" to-port="0" />
+		<edge from-layer="4099" from-port="0" to-layer="4100" to-port="1" />
+		<edge from-layer="4100" from-port="2" to-layer="4109" to-port="0" />
+		<edge from-layer="4101" from-port="0" to-layer="4102" to-port="0" />
+		<edge from-layer="4102" from-port="1" to-layer="4105" to-port="0" />
+		<edge from-layer="4103" from-port="0" to-layer="4104" to-port="0" />
+		<edge from-layer="4104" from-port="1" to-layer="4105" to-port="1" />
+		<edge from-layer="4105" from-port="2" to-layer="4107" to-port="0" />
+		<edge from-layer="4106" from-port="0" to-layer="4107" to-port="1" />
+		<edge from-layer="4107" from-port="2" to-layer="4108" to-port="0" />
+		<edge from-layer="4108" from-port="1" to-layer="4109" to-port="1" />
+		<edge from-layer="4109" from-port="2" to-layer="4112" to-port="0" />
+		<edge from-layer="4110" from-port="0" to-layer="4111" to-port="0" />
+		<edge from-layer="4111" from-port="1" to-layer="4112" to-port="1" />
+		<edge from-layer="4112" from-port="2" to-layer="4113" to-port="0" />
+		<edge from-layer="4113" from-port="1" to-layer="4122" to-port="0" />
+		<edge from-layer="4114" from-port="0" to-layer="4115" to-port="0" />
+		<edge from-layer="4115" from-port="1" to-layer="4118" to-port="0" />
+		<edge from-layer="4116" from-port="0" to-layer="4117" to-port="0" />
+		<edge from-layer="4117" from-port="1" to-layer="4118" to-port="1" />
+		<edge from-layer="4118" from-port="2" to-layer="4120" to-port="0" />
+		<edge from-layer="4119" from-port="0" to-layer="4120" to-port="1" />
+		<edge from-layer="4120" from-port="2" to-layer="4121" to-port="0" />
+		<edge from-layer="4121" from-port="1" to-layer="4122" to-port="1" />
+		<edge from-layer="4122" from-port="2" to-layer="4125" to-port="0" />
+		<edge from-layer="4123" from-port="0" to-layer="4124" to-port="0" />
+		<edge from-layer="4124" from-port="1" to-layer="4125" to-port="1" />
+		<edge from-layer="4125" from-port="2" to-layer="4251" to-port="0" />
+		<edge from-layer="4126" from-port="0" to-layer="4127" to-port="1" />
+		<edge from-layer="4127" from-port="2" to-layer="4129" to-port="0" />
+		<edge from-layer="4128" from-port="0" to-layer="4129" to-port="1" />
+		<edge from-layer="4129" from-port="2" to-layer="4138" to-port="0" />
+		<edge from-layer="4130" from-port="0" to-layer="4131" to-port="0" />
+		<edge from-layer="4131" from-port="1" to-layer="4134" to-port="0" />
+		<edge from-layer="4132" from-port="0" to-layer="4133" to-port="0" />
+		<edge from-layer="4133" from-port="1" to-layer="4134" to-port="1" />
+		<edge from-layer="4134" from-port="2" to-layer="4136" to-port="0" />
+		<edge from-layer="4135" from-port="0" to-layer="4136" to-port="1" />
+		<edge from-layer="4136" from-port="2" to-layer="4137" to-port="0" />
+		<edge from-layer="4137" from-port="1" to-layer="4138" to-port="1" />
+		<edge from-layer="4138" from-port="2" to-layer="4140" to-port="0" />
+		<edge from-layer="4139" from-port="0" to-layer="4140" to-port="1" />
+		<edge from-layer="4140" from-port="2" to-layer="4142" to-port="0" />
+		<edge from-layer="4141" from-port="0" to-layer="4142" to-port="1" />
+		<edge from-layer="4142" from-port="2" to-layer="4223" to-port="0" />
+		<edge from-layer="4142" from-port="2" to-layer="4184" to-port="0" />
+		<edge from-layer="4142" from-port="2" to-layer="4146" to-port="0" />
+		<edge from-layer="4143" from-port="0" to-layer="4146" to-port="1" />
+		<edge from-layer="4144" from-port="0" to-layer="4146" to-port="2" />
 		<edge from-layer="4145" from-port="0" to-layer="4146" to-port="3" />
-		<edge from-layer="4146" from-port="4" to-layer="4150" to-port="0" />
-		<edge from-layer="4147" from-port="0" to-layer="4148" to-port="0" />
-		<edge from-layer="4148" from-port="1" to-layer="4150" to-port="1" />
-		<edge from-layer="4149" from-port="0" to-layer="4150" to-port="2" />
-		<edge from-layer="4150" from-port="3" to-layer="4151" to-port="0" />
-		<edge from-layer="4151" from-port="2" to-layer="4159" to-port="0" />
-		<edge from-layer="4151" from-port="2" to-layer="4152" to-port="0" />
-		<edge from-layer="4152" from-port="1" to-layer="4158" to-port="0" />
-		<edge from-layer="4153" from-port="0" to-layer="4155" to-port="1" />
-		<edge from-layer="4154" from-port="0" to-layer="4155" to-port="2" />
-		<edge from-layer="4155" from-port="3" to-layer="4156" to-port="0" />
-		<edge from-layer="4156" from-port="4" to-layer="4157" to-port="0" />
-		<edge from-layer="4157" from-port="1" to-layer="4158" to-port="1" />
-		<edge from-layer="4158" from-port="2" to-layer="4159" to-port="1" />
-		<edge from-layer="4159" from-port="2" to-layer="4160" to-port="3" />
-		<edge from-layer="4160" from-port="4" to-layer="4162" to-port="0" />
-		<edge from-layer="4161" from-port="0" to-layer="4162" to-port="1" />
-		<edge from-layer="4162" from-port="2" to-layer="4164" to-port="0" />
-		<edge from-layer="4163" from-port="0" to-layer="4164" to-port="1" />
-		<edge from-layer="4164" from-port="2" to-layer="4173" to-port="0" />
-		<edge from-layer="4165" from-port="0" to-layer="4166" to-port="0" />
-		<edge from-layer="4166" from-port="1" to-layer="4169" to-port="0" />
-		<edge from-layer="4167" from-port="0" to-layer="4168" to-port="0" />
-		<edge from-layer="4168" from-port="1" to-layer="4169" to-port="1" />
-		<edge from-layer="4169" from-port="2" to-layer="4171" to-port="0" />
-		<edge from-layer="4170" from-port="0" to-layer="4171" to-port="1" />
-		<edge from-layer="4171" from-port="2" to-layer="4172" to-port="0" />
-		<edge from-layer="4172" from-port="1" to-layer="4173" to-port="1" />
-		<edge from-layer="4173" from-port="2" to-layer="4176" to-port="0" />
-		<edge from-layer="4174" from-port="0" to-layer="4175" to-port="0" />
-		<edge from-layer="4175" from-port="1" to-layer="4176" to-port="1" />
-		<edge from-layer="4176" from-port="2" to-layer="4177" to-port="1" />
+		<edge from-layer="4146" from-port="5" to-layer="4148" to-port="0" />
+		<edge from-layer="4147" from-port="0" to-layer="4148" to-port="1" />
+		<edge from-layer="4148" from-port="2" to-layer="4152" to-port="0" />
+		<edge from-layer="4148" from-port="2" to-layer="4170" to-port="0" />
+		<edge from-layer="4149" from-port="0" to-layer="4152" to-port="1" />
+		<edge from-layer="4150" from-port="0" to-layer="4152" to-port="2" />
+		<edge from-layer="4151" from-port="0" to-layer="4152" to-port="3" />
+		<edge from-layer="4152" from-port="5" to-layer="4163" to-port="0" />
+		<edge from-layer="4152" from-port="5" to-layer="4153" to-port="0" />
+		<edge from-layer="4152" from-port="5" to-layer="4157" to-port="0" />
+		<edge from-layer="4153" from-port="2" to-layer="4166" to-port="0" />
+		<edge from-layer="4154" from-port="0" to-layer="4157" to-port="1" />
+		<edge from-layer="4155" from-port="0" to-layer="4157" to-port="2" />
+		<edge from-layer="4156" from-port="0" to-layer="4157" to-port="3" />
+		<edge from-layer="4157" from-port="5" to-layer="4159" to-port="0" />
+		<edge from-layer="4158" from-port="0" to-layer="4159" to-port="1" />
+		<edge from-layer="4159" from-port="2" to-layer="4164" to-port="0" />
+		<edge from-layer="4160" from-port="0" to-layer="4163" to-port="1" />
+		<edge from-layer="4161" from-port="0" to-layer="4163" to-port="2" />
+		<edge from-layer="4162" from-port="0" to-layer="4163" to-port="3" />
+		<edge from-layer="4163" from-port="5" to-layer="4164" to-port="1" />
+		<edge from-layer="4164" from-port="2" to-layer="4165" to-port="0" />
+		<edge from-layer="4165" from-port="2" to-layer="4166" to-port="1" />
+		<edge from-layer="4166" from-port="2" to-layer="4171" to-port="0" />
+		<edge from-layer="4167" from-port="0" to-layer="4170" to-port="1" />
+		<edge from-layer="4168" from-port="0" to-layer="4170" to-port="2" />
+		<edge from-layer="4169" from-port="0" to-layer="4170" to-port="3" />
+		<edge from-layer="4170" from-port="5" to-layer="4171" to-port="1" />
+		<edge from-layer="4171" from-port="2" to-layer="4235" to-port="0" />
+		<edge from-layer="4172" from-port="0" to-layer="4177" to-port="0" />
+		<edge from-layer="4173" from-port="0" to-layer="4176" to-port="1" />
+		<edge from-layer="4174" from-port="0" to-layer="4176" to-port="2" />
+		<edge from-layer="4175" from-port="0" to-layer="4176" to-port="3" />
+		<edge from-layer="4176" from-port="4" to-layer="4177" to-port="1" />
 		<edge from-layer="4177" from-port="2" to-layer="4178" to-port="0" />
-		<edge from-layer="4178" from-port="2" to-layer="4180" to-port="0" />
-		<edge from-layer="4178" from-port="2" to-layer="4420" to-port="1" />
-		<edge from-layer="4179" from-port="0" to-layer="4180" to-port="1" />
-		<edge from-layer="4180" from-port="2" to-layer="4214" to-port="0" />
-		<edge from-layer="4180" from-port="2" to-layer="4183" to-port="0" />
-		<edge from-layer="4181" from-port="0" to-layer="4182" to-port="0" />
-		<edge from-layer="4182" from-port="1" to-layer="4183" to-port="1" />
-		<edge from-layer="4183" from-port="2" to-layer="4186" to-port="0" />
-		<edge from-layer="4184" from-port="0" to-layer="4185" to-port="0" />
-		<edge from-layer="4185" from-port="1" to-layer="4186" to-port="1" />
-		<edge from-layer="4186" from-port="2" to-layer="4195" to-port="0" />
-		<edge from-layer="4187" from-port="0" to-layer="4188" to-port="0" />
-		<edge from-layer="4188" from-port="1" to-layer="4191" to-port="0" />
-		<edge from-layer="4189" from-port="0" to-layer="4190" to-port="0" />
-		<edge from-layer="4190" from-port="1" to-layer="4191" to-port="1" />
-		<edge from-layer="4191" from-port="2" to-layer="4193" to-port="0" />
-		<edge from-layer="4192" from-port="0" to-layer="4193" to-port="1" />
-		<edge from-layer="4193" from-port="2" to-layer="4194" to-port="0" />
-		<edge from-layer="4194" from-port="1" to-layer="4195" to-port="1" />
-		<edge from-layer="4195" from-port="2" to-layer="4198" to-port="0" />
-		<edge from-layer="4196" from-port="0" to-layer="4197" to-port="0" />
-		<edge from-layer="4197" from-port="1" to-layer="4198" to-port="1" />
-		<edge from-layer="4198" from-port="2" to-layer="4199" to-port="0" />
-		<edge from-layer="4199" from-port="1" to-layer="4208" to-port="0" />
-		<edge from-layer="4200" from-port="0" to-layer="4201" to-port="0" />
-		<edge from-layer="4201" from-port="1" to-layer="4204" to-port="0" />
-		<edge from-layer="4202" from-port="0" to-layer="4203" to-port="0" />
-		<edge from-layer="4203" from-port="1" to-layer="4204" to-port="1" />
-		<edge from-layer="4204" from-port="2" to-layer="4206" to-port="0" />
-		<edge from-layer="4205" from-port="0" to-layer="4206" to-port="1" />
-		<edge from-layer="4206" from-port="2" to-layer="4207" to-port="0" />
-		<edge from-layer="4207" from-port="1" to-layer="4208" to-port="1" />
-		<edge from-layer="4208" from-port="2" to-layer="4211" to-port="0" />
-		<edge from-layer="4209" from-port="0" to-layer="4210" to-port="0" />
-		<edge from-layer="4210" from-port="1" to-layer="4211" to-port="1" />
-		<edge from-layer="4211" from-port="2" to-layer="4419" to-port="0" />
-		<edge from-layer="4212" from-port="0" to-layer="4213" to-port="0" />
-		<edge from-layer="4213" from-port="1" to-layer="4214" to-port="1" />
-		<edge from-layer="4214" from-port="2" to-layer="4217" to-port="0" />
-		<edge from-layer="4215" from-port="0" to-layer="4216" to-port="0" />
-		<edge from-layer="4216" from-port="1" to-layer="4217" to-port="1" />
-		<edge from-layer="4217" from-port="2" to-layer="4226" to-port="0" />
-		<edge from-layer="4218" from-port="0" to-layer="4219" to-port="0" />
-		<edge from-layer="4219" from-port="1" to-layer="4222" to-port="0" />
-		<edge from-layer="4220" from-port="0" to-layer="4221" to-port="0" />
-		<edge from-layer="4221" from-port="1" to-layer="4222" to-port="1" />
-		<edge from-layer="4222" from-port="2" to-layer="4224" to-port="0" />
-		<edge from-layer="4223" from-port="0" to-layer="4224" to-port="1" />
-		<edge from-layer="4224" from-port="2" to-layer="4225" to-port="0" />
-		<edge from-layer="4225" from-port="1" to-layer="4226" to-port="1" />
-		<edge from-layer="4226" from-port="2" to-layer="4229" to-port="0" />
-		<edge from-layer="4227" from-port="0" to-layer="4228" to-port="0" />
-		<edge from-layer="4228" from-port="1" to-layer="4229" to-port="1" />
-		<edge from-layer="4229" from-port="2" to-layer="4231" to-port="0" />
-		<edge from-layer="4230" from-port="0" to-layer="4231" to-port="1" />
-		<edge from-layer="4231" from-port="2" to-layer="4359" to-port="0" />
-		<edge from-layer="4231" from-port="2" to-layer="4235" to-port="0" />
-		<edge from-layer="4231" from-port="2" to-layer="4248" to-port="0" />
-		<edge from-layer="4232" from-port="0" to-layer="4235" to-port="1" />
-		<edge from-layer="4233" from-port="0" to-layer="4235" to-port="2" />
-		<edge from-layer="4234" from-port="0" to-layer="4235" to-port="3" />
+		<edge from-layer="4178" from-port="1" to-layer="4180" to-port="0" />
+		<edge from-layer="4179" from-port="0" to-layer="4180" to-port="2" />
+		<edge from-layer="4180" from-port="3" to-layer="4210" to-port="0" />
+		<edge from-layer="4180" from-port="3" to-layer="4228" to-port="0" />
+		<edge from-layer="4181" from-port="0" to-layer="4184" to-port="1" />
+		<edge from-layer="4182" from-port="0" to-layer="4184" to-port="2" />
+		<edge from-layer="4183" from-port="0" to-layer="4184" to-port="3" />
+		<edge from-layer="4184" from-port="5" to-layer="4186" to-port="0" />
+		<edge from-layer="4185" from-port="0" to-layer="4186" to-port="1" />
+		<edge from-layer="4186" from-port="2" to-layer="4190" to-port="0" />
+		<edge from-layer="4186" from-port="2" to-layer="4208" to-port="0" />
+		<edge from-layer="4187" from-port="0" to-layer="4190" to-port="1" />
+		<edge from-layer="4188" from-port="0" to-layer="4190" to-port="2" />
+		<edge from-layer="4189" from-port="0" to-layer="4190" to-port="3" />
+		<edge from-layer="4190" from-port="5" to-layer="4191" to-port="0" />
+		<edge from-layer="4190" from-port="5" to-layer="4201" to-port="0" />
+		<edge from-layer="4190" from-port="5" to-layer="4195" to-port="0" />
+		<edge from-layer="4191" from-port="2" to-layer="4204" to-port="0" />
+		<edge from-layer="4192" from-port="0" to-layer="4195" to-port="1" />
+		<edge from-layer="4193" from-port="0" to-layer="4195" to-port="2" />
+		<edge from-layer="4194" from-port="0" to-layer="4195" to-port="3" />
+		<edge from-layer="4195" from-port="5" to-layer="4197" to-port="0" />
+		<edge from-layer="4196" from-port="0" to-layer="4197" to-port="1" />
+		<edge from-layer="4197" from-port="2" to-layer="4202" to-port="0" />
+		<edge from-layer="4198" from-port="0" to-layer="4201" to-port="1" />
+		<edge from-layer="4199" from-port="0" to-layer="4201" to-port="2" />
+		<edge from-layer="4200" from-port="0" to-layer="4201" to-port="3" />
+		<edge from-layer="4201" from-port="5" to-layer="4202" to-port="1" />
+		<edge from-layer="4202" from-port="2" to-layer="4203" to-port="0" />
+		<edge from-layer="4203" from-port="2" to-layer="4204" to-port="1" />
+		<edge from-layer="4204" from-port="2" to-layer="4209" to-port="0" />
+		<edge from-layer="4205" from-port="0" to-layer="4208" to-port="1" />
+		<edge from-layer="4206" from-port="0" to-layer="4208" to-port="2" />
+		<edge from-layer="4207" from-port="0" to-layer="4208" to-port="3" />
+		<edge from-layer="4208" from-port="5" to-layer="4209" to-port="1" />
+		<edge from-layer="4209" from-port="2" to-layer="4210" to-port="1" />
+		<edge from-layer="4210" from-port="2" to-layer="4235" to-port="1" />
+		<edge from-layer="4210" from-port="2" to-layer="5123" to-port="0" />
+		<edge from-layer="4211" from-port="0" to-layer="4216" to-port="0" />
+		<edge from-layer="4212" from-port="0" to-layer="4215" to-port="1" />
+		<edge from-layer="4213" from-port="0" to-layer="4215" to-port="2" />
+		<edge from-layer="4214" from-port="0" to-layer="4215" to-port="3" />
+		<edge from-layer="4215" from-port="4" to-layer="4216" to-port="1" />
+		<edge from-layer="4216" from-port="2" to-layer="4217" to-port="0" />
+		<edge from-layer="4217" from-port="1" to-layer="4219" to-port="0" />
+		<edge from-layer="4218" from-port="0" to-layer="4219" to-port="2" />
+		<edge from-layer="4219" from-port="3" to-layer="4226" to-port="0" />
+		<edge from-layer="4220" from-port="0" to-layer="4223" to-port="1" />
+		<edge from-layer="4221" from-port="0" to-layer="4223" to-port="2" />
+		<edge from-layer="4222" from-port="0" to-layer="4223" to-port="3" />
+		<edge from-layer="4223" from-port="5" to-layer="4225" to-port="0" />
+		<edge from-layer="4224" from-port="0" to-layer="4225" to-port="1" />
+		<edge from-layer="4225" from-port="2" to-layer="4226" to-port="1" />
+		<edge from-layer="4226" from-port="2" to-layer="5122" to-port="0" />
+		<edge from-layer="4226" from-port="2" to-layer="4235" to-port="2" />
+		<edge from-layer="4227" from-port="0" to-layer="4234" to-port="1" />
+		<edge from-layer="4228" from-port="1" to-layer="4231" to-port="0" />
+		<edge from-layer="4229" from-port="0" to-layer="4231" to-port="1" />
+		<edge from-layer="4230" from-port="0" to-layer="4231" to-port="2" />
+		<edge from-layer="4231" from-port="3" to-layer="4232" to-port="0" />
+		<edge from-layer="4232" from-port="2" to-layer="4234" to-port="2" />
+		<edge from-layer="4233" from-port="0" to-layer="4234" to-port="3" />
+		<edge from-layer="4234" from-port="5" to-layer="4235" to-port="3" />
 		<edge from-layer="4235" from-port="4" to-layer="4237" to-port="0" />
 		<edge from-layer="4236" from-port="0" to-layer="4237" to-port="1" />
-		<edge from-layer="4237" from-port="2" to-layer="4310" to-port="0" />
-		<edge from-layer="4237" from-port="2" to-layer="4241" to-port="0" />
-		<edge from-layer="4238" from-port="0" to-layer="4241" to-port="1" />
-		<edge from-layer="4239" from-port="0" to-layer="4241" to-port="2" />
-		<edge from-layer="4240" from-port="0" to-layer="4241" to-port="3" />
-		<edge from-layer="4241" from-port="4" to-layer="4275" to-port="0" />
-		<edge from-layer="4241" from-port="4" to-layer="4278" to-port="0" />
-		<edge from-layer="4241" from-port="4" to-layer="4295" to-port="0" />
-		<edge from-layer="4241" from-port="4" to-layer="4289" to-port="0" />
+		<edge from-layer="4237" from-port="2" to-layer="4239" to-port="0" />
+		<edge from-layer="4238" from-port="0" to-layer="4239" to-port="1" />
+		<edge from-layer="4239" from-port="2" to-layer="4248" to-port="0" />
+		<edge from-layer="4240" from-port="0" to-layer="4241" to-port="0" />
+		<edge from-layer="4241" from-port="1" to-layer="4244" to-port="0" />
 		<edge from-layer="4242" from-port="0" to-layer="4243" to-port="0" />
-		<edge from-layer="4243" from-port="1" to-layer="4271" to-port="0" />
-		<edge from-layer="4244" from-port="0" to-layer="4271" to-port="1" />
-		<edge from-layer="4245" from-port="0" to-layer="4248" to-port="1" />
-		<edge from-layer="4246" from-port="0" to-layer="4248" to-port="2" />
-		<edge from-layer="4247" from-port="0" to-layer="4248" to-port="3" />
-		<edge from-layer="4248" from-port="4" to-layer="4250" to-port="0" />
+		<edge from-layer="4243" from-port="1" to-layer="4244" to-port="1" />
+		<edge from-layer="4244" from-port="2" to-layer="4246" to-port="0" />
+		<edge from-layer="4245" from-port="0" to-layer="4246" to-port="1" />
+		<edge from-layer="4246" from-port="2" to-layer="4247" to-port="0" />
+		<edge from-layer="4247" from-port="1" to-layer="4248" to-port="1" />
+		<edge from-layer="4248" from-port="2" to-layer="4250" to-port="0" />
 		<edge from-layer="4249" from-port="0" to-layer="4250" to-port="1" />
-		<edge from-layer="4250" from-port="2" to-layer="4343" to-port="0" />
-		<edge from-layer="4250" from-port="2" to-layer="4251" to-port="0" />
-		<edge from-layer="4250" from-port="2" to-layer="4315" to-port="0" />
-		<edge from-layer="4251" from-port="1" to-layer="4253" to-port="0" />
-		<edge from-layer="4252" from-port="0" to-layer="4253" to-port="2" />
-		<edge from-layer="4253" from-port="3" to-layer="4267" to-port="0" />
-		<edge from-layer="4254" from-port="0" to-layer="4255" to-port="0" />
-		<edge from-layer="4255" from-port="1" to-layer="4260" to-port="0" />
-		<edge from-layer="4256" from-port="0" to-layer="4259" to-port="1" />
-		<edge from-layer="4257" from-port="0" to-layer="4259" to-port="2" />
-		<edge from-layer="4258" from-port="0" to-layer="4259" to-port="3" />
-		<edge from-layer="4259" from-port="4" to-layer="4260" to-port="1" />
-		<edge from-layer="4260" from-port="2" to-layer="4261" to-port="0" />
-		<edge from-layer="4261" from-port="1" to-layer="4263" to-port="0" />
-		<edge from-layer="4262" from-port="0" to-layer="4263" to-port="2" />
-		<edge from-layer="4263" from-port="3" to-layer="4264" to-port="0" />
-		<edge from-layer="4263" from-port="3" to-layer="4345" to-port="0" />
-		<edge from-layer="4264" from-port="1" to-layer="4266" to-port="0" />
-		<edge from-layer="4265" from-port="0" to-layer="4266" to-port="2" />
-		<edge from-layer="4266" from-port="3" to-layer="4267" to-port="1" />
-		<edge from-layer="4267" from-port="2" to-layer="4269" to-port="0" />
-		<edge from-layer="4268" from-port="0" to-layer="4269" to-port="1" />
-		<edge from-layer="4269" from-port="2" to-layer="4271" to-port="2" />
-		<edge from-layer="4269" from-port="2" to-layer="4301" to-port="2" />
-		<edge from-layer="4270" from-port="0" to-layer="4271" to-port="3" />
-		<edge from-layer="4271" from-port="4" to-layer="4273" to-port="0" />
-		<edge from-layer="4272" from-port="0" to-layer="4273" to-port="2" />
-		<edge from-layer="4273" from-port="3" to-layer="4274" to-port="0" />
-		<edge from-layer="4274" from-port="2" to-layer="4316" to-port="1" />
-		<edge from-layer="4274" from-port="2" to-layer="4275" to-port="1" />
-		<edge from-layer="4275" from-port="2" to-layer="4306" to-port="0" />
-		<edge from-layer="4276" from-port="0" to-layer="4286" to-port="0" />
-		<edge from-layer="4277" from-port="0" to-layer="4286" to-port="1" />
-		<edge from-layer="4278" from-port="1" to-layer="4280" to-port="0" />
-		<edge from-layer="4279" from-port="0" to-layer="4280" to-port="2" />
-		<edge from-layer="4280" from-port="3" to-layer="4281" to-port="0" />
-		<edge from-layer="4281" from-port="2" to-layer="4282" to-port="0" />
-		<edge from-layer="4282" from-port="1" to-layer="4284" to-port="0" />
-		<edge from-layer="4283" from-port="0" to-layer="4284" to-port="1" />
-		<edge from-layer="4284" from-port="2" to-layer="4286" to-port="2" />
-		<edge from-layer="4285" from-port="0" to-layer="4286" to-port="3" />
-		<edge from-layer="4286" from-port="4" to-layer="4295" to-port="2" />
-		<edge from-layer="4286" from-port="4" to-layer="4289" to-port="1" />
-		<edge from-layer="4287" from-port="0" to-layer="4289" to-port="2" />
-		<edge from-layer="4288" from-port="0" to-layer="4289" to-port="3" />
-		<edge from-layer="4289" from-port="4" to-layer="4292" to-port="0" />
+		<edge from-layer="4250" from-port="2" to-layer="4251" to-port="1" />
+		<edge from-layer="4251" from-port="2" to-layer="4252" to-port="0" />
+		<edge from-layer="4252" from-port="2" to-layer="4254" to-port="0" />
+		<edge from-layer="4252" from-port="2" to-layer="4410" to-port="1" />
+		<edge from-layer="4253" from-port="0" to-layer="4254" to-port="1" />
+		<edge from-layer="4254" from-port="2" to-layer="4256" to-port="0" />
+		<edge from-layer="4254" from-port="2" to-layer="4285" to-port="0" />
+		<edge from-layer="4255" from-port="0" to-layer="4256" to-port="1" />
+		<edge from-layer="4256" from-port="2" to-layer="4258" to-port="0" />
+		<edge from-layer="4257" from-port="0" to-layer="4258" to-port="1" />
+		<edge from-layer="4258" from-port="2" to-layer="4267" to-port="0" />
+		<edge from-layer="4259" from-port="0" to-layer="4260" to-port="0" />
+		<edge from-layer="4260" from-port="1" to-layer="4263" to-port="0" />
+		<edge from-layer="4261" from-port="0" to-layer="4262" to-port="0" />
+		<edge from-layer="4262" from-port="1" to-layer="4263" to-port="1" />
+		<edge from-layer="4263" from-port="2" to-layer="4265" to-port="0" />
+		<edge from-layer="4264" from-port="0" to-layer="4265" to-port="1" />
+		<edge from-layer="4265" from-port="2" to-layer="4266" to-port="0" />
+		<edge from-layer="4266" from-port="1" to-layer="4267" to-port="1" />
+		<edge from-layer="4267" from-port="2" to-layer="4270" to-port="0" />
+		<edge from-layer="4268" from-port="0" to-layer="4269" to-port="0" />
+		<edge from-layer="4269" from-port="1" to-layer="4270" to-port="1" />
+		<edge from-layer="4270" from-port="2" to-layer="4271" to-port="0" />
+		<edge from-layer="4271" from-port="1" to-layer="4280" to-port="0" />
+		<edge from-layer="4272" from-port="0" to-layer="4273" to-port="0" />
+		<edge from-layer="4273" from-port="1" to-layer="4276" to-port="0" />
+		<edge from-layer="4274" from-port="0" to-layer="4275" to-port="0" />
+		<edge from-layer="4275" from-port="1" to-layer="4276" to-port="1" />
+		<edge from-layer="4276" from-port="2" to-layer="4278" to-port="0" />
+		<edge from-layer="4277" from-port="0" to-layer="4278" to-port="1" />
+		<edge from-layer="4278" from-port="2" to-layer="4279" to-port="0" />
+		<edge from-layer="4279" from-port="1" to-layer="4280" to-port="1" />
+		<edge from-layer="4280" from-port="2" to-layer="4283" to-port="0" />
+		<edge from-layer="4281" from-port="0" to-layer="4282" to-port="0" />
+		<edge from-layer="4282" from-port="1" to-layer="4283" to-port="1" />
+		<edge from-layer="4283" from-port="2" to-layer="4409" to-port="0" />
+		<edge from-layer="4284" from-port="0" to-layer="4285" to-port="1" />
+		<edge from-layer="4285" from-port="2" to-layer="4287" to-port="0" />
+		<edge from-layer="4286" from-port="0" to-layer="4287" to-port="1" />
+		<edge from-layer="4287" from-port="2" to-layer="4296" to-port="0" />
+		<edge from-layer="4288" from-port="0" to-layer="4289" to-port="0" />
+		<edge from-layer="4289" from-port="1" to-layer="4292" to-port="0" />
 		<edge from-layer="4290" from-port="0" to-layer="4291" to-port="0" />
 		<edge from-layer="4291" from-port="1" to-layer="4292" to-port="1" />
-		<edge from-layer="4292" from-port="2" to-layer="4296" to-port="0" />
-		<edge from-layer="4293" from-port="0" to-layer="4295" to-port="1" />
-		<edge from-layer="4294" from-port="0" to-layer="4295" to-port="3" />
-		<edge from-layer="4295" from-port="4" to-layer="4296" to-port="1" />
-		<edge from-layer="4296" from-port="2" to-layer="4305" to-port="0" />
-		<edge from-layer="4297" from-port="0" to-layer="4298" to-port="0" />
-		<edge from-layer="4298" from-port="1" to-layer="4301" to-port="0" />
-		<edge from-layer="4299" from-port="0" to-layer="4301" to-port="1" />
-		<edge from-layer="4300" from-port="0" to-layer="4301" to-port="3" />
-		<edge from-layer="4301" from-port="4" to-layer="4303" to-port="0" />
-		<edge from-layer="4302" from-port="0" to-layer="4303" to-port="2" />
-		<edge from-layer="4303" from-port="3" to-layer="4304" to-port="0" />
-		<edge from-layer="4304" from-port="2" to-layer="4305" to-port="1" />
-		<edge from-layer="4304" from-port="2" to-layer="4338" to-port="1" />
-		<edge from-layer="4305" from-port="2" to-layer="4306" to-port="1" />
-		<edge from-layer="4306" from-port="2" to-layer="4311" to-port="0" />
+		<edge from-layer="4292" from-port="2" to-layer="4294" to-port="0" />
+		<edge from-layer="4293" from-port="0" to-layer="4294" to-port="1" />
+		<edge from-layer="4294" from-port="2" to-layer="4295" to-port="0" />
+		<edge from-layer="4295" from-port="1" to-layer="4296" to-port="1" />
+		<edge from-layer="4296" from-port="2" to-layer="4298" to-port="0" />
+		<edge from-layer="4297" from-port="0" to-layer="4298" to-port="1" />
+		<edge from-layer="4298" from-port="2" to-layer="4300" to-port="0" />
+		<edge from-layer="4299" from-port="0" to-layer="4300" to-port="1" />
+		<edge from-layer="4300" from-port="2" to-layer="4381" to-port="0" />
+		<edge from-layer="4300" from-port="2" to-layer="4304" to-port="0" />
+		<edge from-layer="4300" from-port="2" to-layer="4342" to-port="0" />
+		<edge from-layer="4301" from-port="0" to-layer="4304" to-port="1" />
+		<edge from-layer="4302" from-port="0" to-layer="4304" to-port="2" />
+		<edge from-layer="4303" from-port="0" to-layer="4304" to-port="3" />
+		<edge from-layer="4304" from-port="5" to-layer="4306" to-port="0" />
+		<edge from-layer="4305" from-port="0" to-layer="4306" to-port="1" />
+		<edge from-layer="4306" from-port="2" to-layer="4310" to-port="0" />
+		<edge from-layer="4306" from-port="2" to-layer="4328" to-port="0" />
 		<edge from-layer="4307" from-port="0" to-layer="4310" to-port="1" />
 		<edge from-layer="4308" from-port="0" to-layer="4310" to-port="2" />
 		<edge from-layer="4309" from-port="0" to-layer="4310" to-port="3" />
-		<edge from-layer="4310" from-port="4" to-layer="4311" to-port="1" />
-		<edge from-layer="4311" from-port="2" to-layer="4369" to-port="0" />
-		<edge from-layer="4311" from-port="2" to-layer="4402" to-port="0" />
+		<edge from-layer="4310" from-port="5" to-layer="4311" to-port="0" />
+		<edge from-layer="4310" from-port="5" to-layer="4315" to-port="0" />
+		<edge from-layer="4310" from-port="5" to-layer="4321" to-port="0" />
+		<edge from-layer="4311" from-port="2" to-layer="4324" to-port="0" />
 		<edge from-layer="4312" from-port="0" to-layer="4315" to-port="1" />
 		<edge from-layer="4313" from-port="0" to-layer="4315" to-port="2" />
 		<edge from-layer="4314" from-port="0" to-layer="4315" to-port="3" />
-		<edge from-layer="4315" from-port="4" to-layer="4336" to-port="0" />
-		<edge from-layer="4315" from-port="4" to-layer="4330" to-port="0" />
-		<edge from-layer="4315" from-port="4" to-layer="4316" to-port="0" />
-		<edge from-layer="4315" from-port="4" to-layer="4319" to-port="0" />
-		<edge from-layer="4316" from-port="2" to-layer="4339" to-port="0" />
-		<edge from-layer="4317" from-port="0" to-layer="4327" to-port="0" />
-		<edge from-layer="4318" from-port="0" to-layer="4327" to-port="1" />
-		<edge from-layer="4319" from-port="1" to-layer="4321" to-port="0" />
-		<edge from-layer="4320" from-port="0" to-layer="4321" to-port="2" />
-		<edge from-layer="4321" from-port="3" to-layer="4322" to-port="0" />
+		<edge from-layer="4315" from-port="5" to-layer="4317" to-port="0" />
+		<edge from-layer="4316" from-port="0" to-layer="4317" to-port="1" />
+		<edge from-layer="4317" from-port="2" to-layer="4322" to-port="0" />
+		<edge from-layer="4318" from-port="0" to-layer="4321" to-port="1" />
+		<edge from-layer="4319" from-port="0" to-layer="4321" to-port="2" />
+		<edge from-layer="4320" from-port="0" to-layer="4321" to-port="3" />
+		<edge from-layer="4321" from-port="5" to-layer="4322" to-port="1" />
 		<edge from-layer="4322" from-port="2" to-layer="4323" to-port="0" />
-		<edge from-layer="4323" from-port="1" to-layer="4325" to-port="0" />
-		<edge from-layer="4324" from-port="0" to-layer="4325" to-port="1" />
-		<edge from-layer="4325" from-port="2" to-layer="4327" to-port="2" />
-		<edge from-layer="4326" from-port="0" to-layer="4327" to-port="3" />
-		<edge from-layer="4327" from-port="4" to-layer="4336" to-port="2" />
-		<edge from-layer="4327" from-port="4" to-layer="4330" to-port="1" />
-		<edge from-layer="4328" from-port="0" to-layer="4330" to-port="2" />
-		<edge from-layer="4329" from-port="0" to-layer="4330" to-port="3" />
-		<edge from-layer="4330" from-port="4" to-layer="4333" to-port="0" />
-		<edge from-layer="4331" from-port="0" to-layer="4332" to-port="0" />
-		<edge from-layer="4332" from-port="1" to-layer="4333" to-port="1" />
-		<edge from-layer="4333" from-port="2" to-layer="4337" to-port="0" />
-		<edge from-layer="4334" from-port="0" to-layer="4336" to-port="1" />
-		<edge from-layer="4335" from-port="0" to-layer="4336" to-port="3" />
-		<edge from-layer="4336" from-port="4" to-layer="4337" to-port="1" />
-		<edge from-layer="4337" from-port="2" to-layer="4338" to-port="0" />
-		<edge from-layer="4338" from-port="2" to-layer="4339" to-port="1" />
-		<edge from-layer="4339" from-port="2" to-layer="4344" to-port="0" />
-		<edge from-layer="4340" from-port="0" to-layer="4343" to-port="1" />
-		<edge from-layer="4341" from-port="0" to-layer="4343" to-port="2" />
-		<edge from-layer="4342" from-port="0" to-layer="4343" to-port="3" />
-		<edge from-layer="4343" from-port="4" to-layer="4344" to-port="1" />
-		<edge from-layer="4344" from-port="2" to-layer="4345" to-port="1" />
-		<edge from-layer="4345" from-port="2" to-layer="4366" to-port="0" />
-		<edge from-layer="4345" from-port="2" to-layer="7704" to-port="0" />
-		<edge from-layer="4345" from-port="2" to-layer="4402" to-port="1" />
-		<edge from-layer="4346" from-port="0" to-layer="4347" to-port="0" />
-		<edge from-layer="4347" from-port="1" to-layer="4352" to-port="0" />
-		<edge from-layer="4348" from-port="0" to-layer="4351" to-port="1" />
-		<edge from-layer="4349" from-port="0" to-layer="4351" to-port="2" />
-		<edge from-layer="4350" from-port="0" to-layer="4351" to-port="3" />
-		<edge from-layer="4351" from-port="4" to-layer="4352" to-port="1" />
-		<edge from-layer="4352" from-port="2" to-layer="4353" to-port="0" />
-		<edge from-layer="4353" from-port="1" to-layer="4355" to-port="0" />
-		<edge from-layer="4354" from-port="0" to-layer="4355" to-port="2" />
-		<edge from-layer="4355" from-port="3" to-layer="4362" to-port="0" />
+		<edge from-layer="4323" from-port="2" to-layer="4324" to-port="1" />
+		<edge from-layer="4324" from-port="2" to-layer="4329" to-port="0" />
+		<edge from-layer="4325" from-port="0" to-layer="4328" to-port="1" />
+		<edge from-layer="4326" from-port="0" to-layer="4328" to-port="2" />
+		<edge from-layer="4327" from-port="0" to-layer="4328" to-port="3" />
+		<edge from-layer="4328" from-port="5" to-layer="4329" to-port="1" />
+		<edge from-layer="4329" from-port="2" to-layer="4393" to-port="0" />
+		<edge from-layer="4330" from-port="0" to-layer="4335" to-port="0" />
+		<edge from-layer="4331" from-port="0" to-layer="4334" to-port="1" />
+		<edge from-layer="4332" from-port="0" to-layer="4334" to-port="2" />
+		<edge from-layer="4333" from-port="0" to-layer="4334" to-port="3" />
+		<edge from-layer="4334" from-port="4" to-layer="4335" to-port="1" />
+		<edge from-layer="4335" from-port="2" to-layer="4336" to-port="0" />
+		<edge from-layer="4336" from-port="1" to-layer="4338" to-port="0" />
+		<edge from-layer="4337" from-port="0" to-layer="4338" to-port="2" />
+		<edge from-layer="4338" from-port="3" to-layer="4386" to-port="0" />
+		<edge from-layer="4338" from-port="3" to-layer="4368" to-port="0" />
+		<edge from-layer="4339" from-port="0" to-layer="4342" to-port="1" />
+		<edge from-layer="4340" from-port="0" to-layer="4342" to-port="2" />
+		<edge from-layer="4341" from-port="0" to-layer="4342" to-port="3" />
+		<edge from-layer="4342" from-port="5" to-layer="4344" to-port="0" />
+		<edge from-layer="4343" from-port="0" to-layer="4344" to-port="1" />
+		<edge from-layer="4344" from-port="2" to-layer="4348" to-port="0" />
+		<edge from-layer="4344" from-port="2" to-layer="4366" to-port="0" />
+		<edge from-layer="4345" from-port="0" to-layer="4348" to-port="1" />
+		<edge from-layer="4346" from-port="0" to-layer="4348" to-port="2" />
+		<edge from-layer="4347" from-port="0" to-layer="4348" to-port="3" />
+		<edge from-layer="4348" from-port="5" to-layer="4359" to-port="0" />
+		<edge from-layer="4348" from-port="5" to-layer="4353" to-port="0" />
+		<edge from-layer="4348" from-port="5" to-layer="4349" to-port="0" />
+		<edge from-layer="4349" from-port="2" to-layer="4362" to-port="0" />
+		<edge from-layer="4350" from-port="0" to-layer="4353" to-port="1" />
+		<edge from-layer="4351" from-port="0" to-layer="4353" to-port="2" />
+		<edge from-layer="4352" from-port="0" to-layer="4353" to-port="3" />
+		<edge from-layer="4353" from-port="5" to-layer="4355" to-port="0" />
+		<edge from-layer="4354" from-port="0" to-layer="4355" to-port="1" />
+		<edge from-layer="4355" from-port="2" to-layer="4360" to-port="0" />
 		<edge from-layer="4356" from-port="0" to-layer="4359" to-port="1" />
 		<edge from-layer="4357" from-port="0" to-layer="4359" to-port="2" />
 		<edge from-layer="4358" from-port="0" to-layer="4359" to-port="3" />
-		<edge from-layer="4359" from-port="4" to-layer="4361" to-port="0" />
-		<edge from-layer="4360" from-port="0" to-layer="4361" to-port="1" />
+		<edge from-layer="4359" from-port="5" to-layer="4360" to-port="1" />
+		<edge from-layer="4360" from-port="2" to-layer="4361" to-port="0" />
 		<edge from-layer="4361" from-port="2" to-layer="4362" to-port="1" />
-		<edge from-layer="4362" from-port="2" to-layer="7703" to-port="0" />
-		<edge from-layer="4362" from-port="2" to-layer="4402" to-port="2" />
-		<edge from-layer="4363" from-port="0" to-layer="4381" to-port="0" />
-		<edge from-layer="4364" from-port="0" to-layer="4376" to-port="0" />
-		<edge from-layer="4365" from-port="0" to-layer="4376" to-port="1" />
-		<edge from-layer="4365" from-port="0" to-layer="4379" to-port="1" />
-		<edge from-layer="4366" from-port="1" to-layer="4368" to-port="0" />
-		<edge from-layer="4367" from-port="0" to-layer="4368" to-port="2" />
-		<edge from-layer="4368" from-port="3" to-layer="4372" to-port="0" />
-		<edge from-layer="4368" from-port="3" to-layer="4378" to-port="0" />
-		<edge from-layer="4369" from-port="1" to-layer="4371" to-port="0" />
-		<edge from-layer="4369" from-port="1" to-layer="4397" to-port="0" />
-		<edge from-layer="4370" from-port="0" to-layer="4371" to-port="2" />
-		<edge from-layer="4371" from-port="3" to-layer="4372" to-port="1" />
-		<edge from-layer="4372" from-port="2" to-layer="4374" to-port="0" />
-		<edge from-layer="4373" from-port="0" to-layer="4374" to-port="1" />
-		<edge from-layer="4373" from-port="0" to-layer="4378" to-port="1" />
-		<edge from-layer="4374" from-port="2" to-layer="4376" to-port="2" />
-		<edge from-layer="4375" from-port="0" to-layer="4376" to-port="3" />
-		<edge from-layer="4375" from-port="0" to-layer="4379" to-port="3" />
-		<edge from-layer="4376" from-port="4" to-layer="4381" to-port="1" />
-		<edge from-layer="4377" from-port="0" to-layer="4379" to-port="0" />
-		<edge from-layer="4378" from-port="2" to-layer="4386" to-port="2" />
-		<edge from-layer="4378" from-port="2" to-layer="4379" to-port="2" />
-		<edge from-layer="4379" from-port="4" to-layer="4381" to-port="2" />
+		<edge from-layer="4362" from-port="2" to-layer="4367" to-port="0" />
+		<edge from-layer="4363" from-port="0" to-layer="4366" to-port="1" />
+		<edge from-layer="4364" from-port="0" to-layer="4366" to-port="2" />
+		<edge from-layer="4365" from-port="0" to-layer="4366" to-port="3" />
+		<edge from-layer="4366" from-port="5" to-layer="4367" to-port="1" />
+		<edge from-layer="4367" from-port="2" to-layer="4368" to-port="1" />
+		<edge from-layer="4368" from-port="2" to-layer="5121" to-port="0" />
+		<edge from-layer="4368" from-port="2" to-layer="4393" to-port="1" />
+		<edge from-layer="4369" from-port="0" to-layer="4374" to-port="0" />
+		<edge from-layer="4370" from-port="0" to-layer="4373" to-port="1" />
+		<edge from-layer="4371" from-port="0" to-layer="4373" to-port="2" />
+		<edge from-layer="4372" from-port="0" to-layer="4373" to-port="3" />
+		<edge from-layer="4373" from-port="4" to-layer="4374" to-port="1" />
+		<edge from-layer="4374" from-port="2" to-layer="4375" to-port="0" />
+		<edge from-layer="4375" from-port="1" to-layer="4377" to-port="0" />
+		<edge from-layer="4376" from-port="0" to-layer="4377" to-port="2" />
+		<edge from-layer="4377" from-port="3" to-layer="4384" to-port="0" />
+		<edge from-layer="4378" from-port="0" to-layer="4381" to-port="1" />
+		<edge from-layer="4379" from-port="0" to-layer="4381" to-port="2" />
 		<edge from-layer="4380" from-port="0" to-layer="4381" to-port="3" />
-		<edge from-layer="4381" from-port="4" to-layer="4388" to-port="0" />
-		<edge from-layer="4382" from-port="0" to-layer="4388" to-port="1" />
-		<edge from-layer="4383" from-port="0" to-layer="4386" to-port="0" />
-		<edge from-layer="4384" from-port="0" to-layer="4386" to-port="1" />
-		<edge from-layer="4385" from-port="0" to-layer="4386" to-port="3" />
-		<edge from-layer="4386" from-port="4" to-layer="4388" to-port="2" />
-		<edge from-layer="4387" from-port="0" to-layer="4388" to-port="3" />
-		<edge from-layer="4388" from-port="4" to-layer="4392" to-port="0" />
-		<edge from-layer="4389" from-port="0" to-layer="4390" to-port="0" />
-		<edge from-layer="4390" from-port="1" to-layer="4392" to-port="1" />
-		<edge from-layer="4391" from-port="0" to-layer="4392" to-port="2" />
-		<edge from-layer="4392" from-port="3" to-layer="4393" to-port="0" />
-		<edge from-layer="4393" from-port="2" to-layer="4401" to-port="0" />
-		<edge from-layer="4393" from-port="2" to-layer="4394" to-port="0" />
-		<edge from-layer="4394" from-port="1" to-layer="4400" to-port="0" />
-		<edge from-layer="4395" from-port="0" to-layer="4397" to-port="1" />
-		<edge from-layer="4396" from-port="0" to-layer="4397" to-port="2" />
-		<edge from-layer="4397" from-port="3" to-layer="4398" to-port="0" />
-		<edge from-layer="4398" from-port="4" to-layer="4399" to-port="0" />
-		<edge from-layer="4399" from-port="1" to-layer="4400" to-port="1" />
-		<edge from-layer="4400" from-port="2" to-layer="4401" to-port="1" />
-		<edge from-layer="4401" from-port="2" to-layer="4402" to-port="3" />
-		<edge from-layer="4402" from-port="4" to-layer="4404" to-port="0" />
+		<edge from-layer="4381" from-port="5" to-layer="4383" to-port="0" />
+		<edge from-layer="4382" from-port="0" to-layer="4383" to-port="1" />
+		<edge from-layer="4383" from-port="2" to-layer="4384" to-port="1" />
+		<edge from-layer="4384" from-port="2" to-layer="5120" to-port="0" />
+		<edge from-layer="4384" from-port="2" to-layer="4393" to-port="2" />
+		<edge from-layer="4385" from-port="0" to-layer="4392" to-port="1" />
+		<edge from-layer="4386" from-port="1" to-layer="4389" to-port="0" />
+		<edge from-layer="4387" from-port="0" to-layer="4389" to-port="1" />
+		<edge from-layer="4388" from-port="0" to-layer="4389" to-port="2" />
+		<edge from-layer="4389" from-port="3" to-layer="4390" to-port="0" />
+		<edge from-layer="4390" from-port="2" to-layer="4392" to-port="2" />
+		<edge from-layer="4391" from-port="0" to-layer="4392" to-port="3" />
+		<edge from-layer="4392" from-port="5" to-layer="4393" to-port="3" />
+		<edge from-layer="4393" from-port="4" to-layer="4395" to-port="0" />
+		<edge from-layer="4394" from-port="0" to-layer="4395" to-port="1" />
+		<edge from-layer="4395" from-port="2" to-layer="4397" to-port="0" />
+		<edge from-layer="4396" from-port="0" to-layer="4397" to-port="1" />
+		<edge from-layer="4397" from-port="2" to-layer="4406" to-port="0" />
+		<edge from-layer="4398" from-port="0" to-layer="4399" to-port="0" />
+		<edge from-layer="4399" from-port="1" to-layer="4402" to-port="0" />
+		<edge from-layer="4400" from-port="0" to-layer="4401" to-port="0" />
+		<edge from-layer="4401" from-port="1" to-layer="4402" to-port="1" />
+		<edge from-layer="4402" from-port="2" to-layer="4404" to-port="0" />
 		<edge from-layer="4403" from-port="0" to-layer="4404" to-port="1" />
-		<edge from-layer="4404" from-port="2" to-layer="4406" to-port="0" />
-		<edge from-layer="4405" from-port="0" to-layer="4406" to-port="1" />
-		<edge from-layer="4406" from-port="2" to-layer="4415" to-port="0" />
-		<edge from-layer="4407" from-port="0" to-layer="4408" to-port="0" />
-		<edge from-layer="4408" from-port="1" to-layer="4411" to-port="0" />
-		<edge from-layer="4409" from-port="0" to-layer="4410" to-port="0" />
-		<edge from-layer="4410" from-port="1" to-layer="4411" to-port="1" />
-		<edge from-layer="4411" from-port="2" to-layer="4413" to-port="0" />
-		<edge from-layer="4412" from-port="0" to-layer="4413" to-port="1" />
-		<edge from-layer="4413" from-port="2" to-layer="4414" to-port="0" />
-		<edge from-layer="4414" from-port="1" to-layer="4415" to-port="1" />
-		<edge from-layer="4415" from-port="2" to-layer="4418" to-port="0" />
-		<edge from-layer="4416" from-port="0" to-layer="4417" to-port="0" />
-		<edge from-layer="4417" from-port="1" to-layer="4418" to-port="1" />
-		<edge from-layer="4418" from-port="2" to-layer="4419" to-port="1" />
-		<edge from-layer="4419" from-port="2" to-layer="4420" to-port="0" />
-		<edge from-layer="4420" from-port="2" to-layer="4662" to-port="1" />
-		<edge from-layer="4420" from-port="2" to-layer="4422" to-port="0" />
-		<edge from-layer="4421" from-port="0" to-layer="4422" to-port="1" />
-		<edge from-layer="4422" from-port="2" to-layer="4456" to-port="0" />
-		<edge from-layer="4422" from-port="2" to-layer="4425" to-port="0" />
-		<edge from-layer="4423" from-port="0" to-layer="4424" to-port="0" />
+		<edge from-layer="4404" from-port="2" to-layer="4405" to-port="0" />
+		<edge from-layer="4405" from-port="1" to-layer="4406" to-port="1" />
+		<edge from-layer="4406" from-port="2" to-layer="4408" to-port="0" />
+		<edge from-layer="4407" from-port="0" to-layer="4408" to-port="1" />
+		<edge from-layer="4408" from-port="2" to-layer="4409" to-port="1" />
+		<edge from-layer="4409" from-port="2" to-layer="4410" to-port="0" />
+		<edge from-layer="4410" from-port="2" to-layer="4568" to-port="1" />
+		<edge from-layer="4410" from-port="2" to-layer="4412" to-port="0" />
+		<edge from-layer="4411" from-port="0" to-layer="4412" to-port="1" />
+		<edge from-layer="4412" from-port="2" to-layer="4414" to-port="0" />
+		<edge from-layer="4412" from-port="2" to-layer="4443" to-port="0" />
+		<edge from-layer="4413" from-port="0" to-layer="4414" to-port="1" />
+		<edge from-layer="4414" from-port="2" to-layer="4416" to-port="0" />
+		<edge from-layer="4415" from-port="0" to-layer="4416" to-port="1" />
+		<edge from-layer="4416" from-port="2" to-layer="4425" to-port="0" />
+		<edge from-layer="4417" from-port="0" to-layer="4418" to-port="0" />
+		<edge from-layer="4418" from-port="1" to-layer="4421" to-port="0" />
+		<edge from-layer="4419" from-port="0" to-layer="4420" to-port="0" />
+		<edge from-layer="4420" from-port="1" to-layer="4421" to-port="1" />
+		<edge from-layer="4421" from-port="2" to-layer="4423" to-port="0" />
+		<edge from-layer="4422" from-port="0" to-layer="4423" to-port="1" />
+		<edge from-layer="4423" from-port="2" to-layer="4424" to-port="0" />
 		<edge from-layer="4424" from-port="1" to-layer="4425" to-port="1" />
 		<edge from-layer="4425" from-port="2" to-layer="4428" to-port="0" />
 		<edge from-layer="4426" from-port="0" to-layer="4427" to-port="0" />
 		<edge from-layer="4427" from-port="1" to-layer="4428" to-port="1" />
-		<edge from-layer="4428" from-port="2" to-layer="4437" to-port="0" />
-		<edge from-layer="4429" from-port="0" to-layer="4430" to-port="0" />
-		<edge from-layer="4430" from-port="1" to-layer="4433" to-port="0" />
-		<edge from-layer="4431" from-port="0" to-layer="4432" to-port="0" />
-		<edge from-layer="4432" from-port="1" to-layer="4433" to-port="1" />
-		<edge from-layer="4433" from-port="2" to-layer="4435" to-port="0" />
-		<edge from-layer="4434" from-port="0" to-layer="4435" to-port="1" />
-		<edge from-layer="4435" from-port="2" to-layer="4436" to-port="0" />
-		<edge from-layer="4436" from-port="1" to-layer="4437" to-port="1" />
-		<edge from-layer="4437" from-port="2" to-layer="4440" to-port="0" />
-		<edge from-layer="4438" from-port="0" to-layer="4439" to-port="0" />
-		<edge from-layer="4439" from-port="1" to-layer="4440" to-port="1" />
-		<edge from-layer="4440" from-port="2" to-layer="4441" to-port="0" />
-		<edge from-layer="4441" from-port="1" to-layer="4450" to-port="0" />
-		<edge from-layer="4442" from-port="0" to-layer="4443" to-port="0" />
-		<edge from-layer="4443" from-port="1" to-layer="4446" to-port="0" />
-		<edge from-layer="4444" from-port="0" to-layer="4445" to-port="0" />
-		<edge from-layer="4445" from-port="1" to-layer="4446" to-port="1" />
-		<edge from-layer="4446" from-port="2" to-layer="4448" to-port="0" />
-		<edge from-layer="4447" from-port="0" to-layer="4448" to-port="1" />
-		<edge from-layer="4448" from-port="2" to-layer="4449" to-port="0" />
+		<edge from-layer="4428" from-port="2" to-layer="4429" to-port="0" />
+		<edge from-layer="4429" from-port="1" to-layer="4438" to-port="0" />
+		<edge from-layer="4430" from-port="0" to-layer="4431" to-port="0" />
+		<edge from-layer="4431" from-port="1" to-layer="4434" to-port="0" />
+		<edge from-layer="4432" from-port="0" to-layer="4433" to-port="0" />
+		<edge from-layer="4433" from-port="1" to-layer="4434" to-port="1" />
+		<edge from-layer="4434" from-port="2" to-layer="4436" to-port="0" />
+		<edge from-layer="4435" from-port="0" to-layer="4436" to-port="1" />
+		<edge from-layer="4436" from-port="2" to-layer="4437" to-port="0" />
+		<edge from-layer="4437" from-port="1" to-layer="4438" to-port="1" />
+		<edge from-layer="4438" from-port="2" to-layer="4441" to-port="0" />
+		<edge from-layer="4439" from-port="0" to-layer="4440" to-port="0" />
+		<edge from-layer="4440" from-port="1" to-layer="4441" to-port="1" />
+		<edge from-layer="4441" from-port="2" to-layer="4567" to-port="0" />
+		<edge from-layer="4442" from-port="0" to-layer="4443" to-port="1" />
+		<edge from-layer="4443" from-port="2" to-layer="4445" to-port="0" />
+		<edge from-layer="4444" from-port="0" to-layer="4445" to-port="1" />
+		<edge from-layer="4445" from-port="2" to-layer="4454" to-port="0" />
+		<edge from-layer="4446" from-port="0" to-layer="4447" to-port="0" />
+		<edge from-layer="4447" from-port="1" to-layer="4450" to-port="0" />
+		<edge from-layer="4448" from-port="0" to-layer="4449" to-port="0" />
 		<edge from-layer="4449" from-port="1" to-layer="4450" to-port="1" />
-		<edge from-layer="4450" from-port="2" to-layer="4453" to-port="0" />
-		<edge from-layer="4451" from-port="0" to-layer="4452" to-port="0" />
-		<edge from-layer="4452" from-port="1" to-layer="4453" to-port="1" />
-		<edge from-layer="4453" from-port="2" to-layer="4661" to-port="0" />
-		<edge from-layer="4454" from-port="0" to-layer="4455" to-port="0" />
-		<edge from-layer="4455" from-port="1" to-layer="4456" to-port="1" />
-		<edge from-layer="4456" from-port="2" to-layer="4459" to-port="0" />
-		<edge from-layer="4457" from-port="0" to-layer="4458" to-port="0" />
-		<edge from-layer="4458" from-port="1" to-layer="4459" to-port="1" />
-		<edge from-layer="4459" from-port="2" to-layer="4468" to-port="0" />
-		<edge from-layer="4460" from-port="0" to-layer="4461" to-port="0" />
-		<edge from-layer="4461" from-port="1" to-layer="4464" to-port="0" />
-		<edge from-layer="4462" from-port="0" to-layer="4463" to-port="0" />
-		<edge from-layer="4463" from-port="1" to-layer="4464" to-port="1" />
-		<edge from-layer="4464" from-port="2" to-layer="4466" to-port="0" />
-		<edge from-layer="4465" from-port="0" to-layer="4466" to-port="1" />
-		<edge from-layer="4466" from-port="2" to-layer="4467" to-port="0" />
-		<edge from-layer="4467" from-port="1" to-layer="4468" to-port="1" />
-		<edge from-layer="4468" from-port="2" to-layer="4471" to-port="0" />
-		<edge from-layer="4469" from-port="0" to-layer="4470" to-port="0" />
-		<edge from-layer="4470" from-port="1" to-layer="4471" to-port="1" />
-		<edge from-layer="4471" from-port="2" to-layer="4473" to-port="0" />
-		<edge from-layer="4472" from-port="0" to-layer="4473" to-port="1" />
-		<edge from-layer="4473" from-port="2" to-layer="4601" to-port="0" />
-		<edge from-layer="4473" from-port="2" to-layer="4477" to-port="0" />
-		<edge from-layer="4473" from-port="2" to-layer="4490" to-port="0" />
-		<edge from-layer="4474" from-port="0" to-layer="4477" to-port="1" />
-		<edge from-layer="4475" from-port="0" to-layer="4477" to-port="2" />
-		<edge from-layer="4476" from-port="0" to-layer="4477" to-port="3" />
-		<edge from-layer="4477" from-port="4" to-layer="4479" to-port="0" />
-		<edge from-layer="4478" from-port="0" to-layer="4479" to-port="1" />
-		<edge from-layer="4479" from-port="2" to-layer="4483" to-port="0" />
-		<edge from-layer="4479" from-port="2" to-layer="4552" to-port="0" />
-		<edge from-layer="4480" from-port="0" to-layer="4483" to-port="1" />
-		<edge from-layer="4481" from-port="0" to-layer="4483" to-port="2" />
-		<edge from-layer="4482" from-port="0" to-layer="4483" to-port="3" />
-		<edge from-layer="4483" from-port="4" to-layer="4531" to-port="0" />
-		<edge from-layer="4483" from-port="4" to-layer="4517" to-port="0" />
-		<edge from-layer="4483" from-port="4" to-layer="4520" to-port="0" />
-		<edge from-layer="4483" from-port="4" to-layer="4537" to-port="0" />
-		<edge from-layer="4484" from-port="0" to-layer="4485" to-port="0" />
-		<edge from-layer="4485" from-port="1" to-layer="4513" to-port="0" />
-		<edge from-layer="4486" from-port="0" to-layer="4513" to-port="1" />
-		<edge from-layer="4487" from-port="0" to-layer="4490" to-port="1" />
-		<edge from-layer="4488" from-port="0" to-layer="4490" to-port="2" />
-		<edge from-layer="4489" from-port="0" to-layer="4490" to-port="3" />
-		<edge from-layer="4490" from-port="4" to-layer="4492" to-port="0" />
-		<edge from-layer="4491" from-port="0" to-layer="4492" to-port="1" />
-		<edge from-layer="4492" from-port="2" to-layer="4493" to-port="0" />
-		<edge from-layer="4492" from-port="2" to-layer="4585" to-port="0" />
-		<edge from-layer="4492" from-port="2" to-layer="4557" to-port="0" />
-		<edge from-layer="4493" from-port="1" to-layer="4495" to-port="0" />
-		<edge from-layer="4494" from-port="0" to-layer="4495" to-port="2" />
-		<edge from-layer="4495" from-port="3" to-layer="4509" to-port="0" />
-		<edge from-layer="4496" from-port="0" to-layer="4497" to-port="0" />
-		<edge from-layer="4497" from-port="1" to-layer="4502" to-port="0" />
-		<edge from-layer="4498" from-port="0" to-layer="4501" to-port="1" />
-		<edge from-layer="4499" from-port="0" to-layer="4501" to-port="2" />
-		<edge from-layer="4500" from-port="0" to-layer="4501" to-port="3" />
-		<edge from-layer="4501" from-port="4" to-layer="4502" to-port="1" />
-		<edge from-layer="4502" from-port="2" to-layer="4503" to-port="0" />
-		<edge from-layer="4503" from-port="1" to-layer="4505" to-port="0" />
-		<edge from-layer="4504" from-port="0" to-layer="4505" to-port="2" />
-		<edge from-layer="4505" from-port="3" to-layer="4506" to-port="0" />
-		<edge from-layer="4505" from-port="3" to-layer="4587" to-port="0" />
-		<edge from-layer="4506" from-port="1" to-layer="4508" to-port="0" />
-		<edge from-layer="4507" from-port="0" to-layer="4508" to-port="2" />
-		<edge from-layer="4508" from-port="3" to-layer="4509" to-port="1" />
-		<edge from-layer="4509" from-port="2" to-layer="4511" to-port="0" />
-		<edge from-layer="4510" from-port="0" to-layer="4511" to-port="1" />
-		<edge from-layer="4511" from-port="2" to-layer="4513" to-port="2" />
-		<edge from-layer="4511" from-port="2" to-layer="4543" to-port="2" />
-		<edge from-layer="4512" from-port="0" to-layer="4513" to-port="3" />
-		<edge from-layer="4513" from-port="4" to-layer="4515" to-port="0" />
-		<edge from-layer="4514" from-port="0" to-layer="4515" to-port="2" />
-		<edge from-layer="4515" from-port="3" to-layer="4516" to-port="0" />
-		<edge from-layer="4516" from-port="2" to-layer="4517" to-port="1" />
-		<edge from-layer="4516" from-port="2" to-layer="4558" to-port="1" />
-		<edge from-layer="4517" from-port="2" to-layer="4548" to-port="0" />
-		<edge from-layer="4518" from-port="0" to-layer="4528" to-port="0" />
-		<edge from-layer="4519" from-port="0" to-layer="4528" to-port="1" />
-		<edge from-layer="4520" from-port="1" to-layer="4522" to-port="0" />
-		<edge from-layer="4521" from-port="0" to-layer="4522" to-port="2" />
-		<edge from-layer="4522" from-port="3" to-layer="4523" to-port="0" />
-		<edge from-layer="4523" from-port="2" to-layer="4524" to-port="0" />
-		<edge from-layer="4524" from-port="1" to-layer="4526" to-port="0" />
-		<edge from-layer="4525" from-port="0" to-layer="4526" to-port="1" />
-		<edge from-layer="4526" from-port="2" to-layer="4528" to-port="2" />
-		<edge from-layer="4527" from-port="0" to-layer="4528" to-port="3" />
-		<edge from-layer="4528" from-port="4" to-layer="4537" to-port="2" />
-		<edge from-layer="4528" from-port="4" to-layer="4531" to-port="1" />
+		<edge from-layer="4450" from-port="2" to-layer="4452" to-port="0" />
+		<edge from-layer="4451" from-port="0" to-layer="4452" to-port="1" />
+		<edge from-layer="4452" from-port="2" to-layer="4453" to-port="0" />
+		<edge from-layer="4453" from-port="1" to-layer="4454" to-port="1" />
+		<edge from-layer="4454" from-port="2" to-layer="4456" to-port="0" />
+		<edge from-layer="4455" from-port="0" to-layer="4456" to-port="1" />
+		<edge from-layer="4456" from-port="2" to-layer="4458" to-port="0" />
+		<edge from-layer="4457" from-port="0" to-layer="4458" to-port="1" />
+		<edge from-layer="4458" from-port="2" to-layer="4500" to-port="0" />
+		<edge from-layer="4458" from-port="2" to-layer="4462" to-port="0" />
+		<edge from-layer="4458" from-port="2" to-layer="4539" to-port="0" />
+		<edge from-layer="4459" from-port="0" to-layer="4462" to-port="1" />
+		<edge from-layer="4460" from-port="0" to-layer="4462" to-port="2" />
+		<edge from-layer="4461" from-port="0" to-layer="4462" to-port="3" />
+		<edge from-layer="4462" from-port="5" to-layer="4464" to-port="0" />
+		<edge from-layer="4463" from-port="0" to-layer="4464" to-port="1" />
+		<edge from-layer="4464" from-port="2" to-layer="4486" to-port="0" />
+		<edge from-layer="4464" from-port="2" to-layer="4468" to-port="0" />
+		<edge from-layer="4465" from-port="0" to-layer="4468" to-port="1" />
+		<edge from-layer="4466" from-port="0" to-layer="4468" to-port="2" />
+		<edge from-layer="4467" from-port="0" to-layer="4468" to-port="3" />
+		<edge from-layer="4468" from-port="5" to-layer="4473" to-port="0" />
+		<edge from-layer="4468" from-port="5" to-layer="4469" to-port="0" />
+		<edge from-layer="4468" from-port="5" to-layer="4479" to-port="0" />
+		<edge from-layer="4469" from-port="2" to-layer="4482" to-port="0" />
+		<edge from-layer="4470" from-port="0" to-layer="4473" to-port="1" />
+		<edge from-layer="4471" from-port="0" to-layer="4473" to-port="2" />
+		<edge from-layer="4472" from-port="0" to-layer="4473" to-port="3" />
+		<edge from-layer="4473" from-port="5" to-layer="4475" to-port="0" />
+		<edge from-layer="4474" from-port="0" to-layer="4475" to-port="1" />
+		<edge from-layer="4475" from-port="2" to-layer="4480" to-port="0" />
+		<edge from-layer="4476" from-port="0" to-layer="4479" to-port="1" />
+		<edge from-layer="4477" from-port="0" to-layer="4479" to-port="2" />
+		<edge from-layer="4478" from-port="0" to-layer="4479" to-port="3" />
+		<edge from-layer="4479" from-port="5" to-layer="4480" to-port="1" />
+		<edge from-layer="4480" from-port="2" to-layer="4481" to-port="0" />
+		<edge from-layer="4481" from-port="2" to-layer="4482" to-port="1" />
+		<edge from-layer="4482" from-port="2" to-layer="4487" to-port="0" />
+		<edge from-layer="4483" from-port="0" to-layer="4486" to-port="1" />
+		<edge from-layer="4484" from-port="0" to-layer="4486" to-port="2" />
+		<edge from-layer="4485" from-port="0" to-layer="4486" to-port="3" />
+		<edge from-layer="4486" from-port="5" to-layer="4487" to-port="1" />
+		<edge from-layer="4487" from-port="2" to-layer="4551" to-port="0" />
+		<edge from-layer="4488" from-port="0" to-layer="4493" to-port="0" />
+		<edge from-layer="4489" from-port="0" to-layer="4492" to-port="1" />
+		<edge from-layer="4490" from-port="0" to-layer="4492" to-port="2" />
+		<edge from-layer="4491" from-port="0" to-layer="4492" to-port="3" />
+		<edge from-layer="4492" from-port="4" to-layer="4493" to-port="1" />
+		<edge from-layer="4493" from-port="2" to-layer="4494" to-port="0" />
+		<edge from-layer="4494" from-port="1" to-layer="4496" to-port="0" />
+		<edge from-layer="4495" from-port="0" to-layer="4496" to-port="2" />
+		<edge from-layer="4496" from-port="3" to-layer="4544" to-port="0" />
+		<edge from-layer="4496" from-port="3" to-layer="4526" to-port="0" />
+		<edge from-layer="4497" from-port="0" to-layer="4500" to-port="1" />
+		<edge from-layer="4498" from-port="0" to-layer="4500" to-port="2" />
+		<edge from-layer="4499" from-port="0" to-layer="4500" to-port="3" />
+		<edge from-layer="4500" from-port="5" to-layer="4502" to-port="0" />
+		<edge from-layer="4501" from-port="0" to-layer="4502" to-port="1" />
+		<edge from-layer="4502" from-port="2" to-layer="4506" to-port="0" />
+		<edge from-layer="4502" from-port="2" to-layer="4524" to-port="0" />
+		<edge from-layer="4503" from-port="0" to-layer="4506" to-port="1" />
+		<edge from-layer="4504" from-port="0" to-layer="4506" to-port="2" />
+		<edge from-layer="4505" from-port="0" to-layer="4506" to-port="3" />
+		<edge from-layer="4506" from-port="5" to-layer="4507" to-port="0" />
+		<edge from-layer="4506" from-port="5" to-layer="4511" to-port="0" />
+		<edge from-layer="4506" from-port="5" to-layer="4517" to-port="0" />
+		<edge from-layer="4507" from-port="2" to-layer="4520" to-port="0" />
+		<edge from-layer="4508" from-port="0" to-layer="4511" to-port="1" />
+		<edge from-layer="4509" from-port="0" to-layer="4511" to-port="2" />
+		<edge from-layer="4510" from-port="0" to-layer="4511" to-port="3" />
+		<edge from-layer="4511" from-port="5" to-layer="4513" to-port="0" />
+		<edge from-layer="4512" from-port="0" to-layer="4513" to-port="1" />
+		<edge from-layer="4513" from-port="2" to-layer="4518" to-port="0" />
+		<edge from-layer="4514" from-port="0" to-layer="4517" to-port="1" />
+		<edge from-layer="4515" from-port="0" to-layer="4517" to-port="2" />
+		<edge from-layer="4516" from-port="0" to-layer="4517" to-port="3" />
+		<edge from-layer="4517" from-port="5" to-layer="4518" to-port="1" />
+		<edge from-layer="4518" from-port="2" to-layer="4519" to-port="0" />
+		<edge from-layer="4519" from-port="2" to-layer="4520" to-port="1" />
+		<edge from-layer="4520" from-port="2" to-layer="4525" to-port="0" />
+		<edge from-layer="4521" from-port="0" to-layer="4524" to-port="1" />
+		<edge from-layer="4522" from-port="0" to-layer="4524" to-port="2" />
+		<edge from-layer="4523" from-port="0" to-layer="4524" to-port="3" />
+		<edge from-layer="4524" from-port="5" to-layer="4525" to-port="1" />
+		<edge from-layer="4525" from-port="2" to-layer="4526" to-port="1" />
+		<edge from-layer="4526" from-port="2" to-layer="5119" to-port="0" />
+		<edge from-layer="4526" from-port="2" to-layer="4551" to-port="1" />
+		<edge from-layer="4527" from-port="0" to-layer="4532" to-port="0" />
+		<edge from-layer="4528" from-port="0" to-layer="4531" to-port="1" />
 		<edge from-layer="4529" from-port="0" to-layer="4531" to-port="2" />
 		<edge from-layer="4530" from-port="0" to-layer="4531" to-port="3" />
-		<edge from-layer="4531" from-port="4" to-layer="4534" to-port="0" />
-		<edge from-layer="4532" from-port="0" to-layer="4533" to-port="0" />
-		<edge from-layer="4533" from-port="1" to-layer="4534" to-port="1" />
-		<edge from-layer="4534" from-port="2" to-layer="4538" to-port="0" />
-		<edge from-layer="4535" from-port="0" to-layer="4537" to-port="1" />
-		<edge from-layer="4536" from-port="0" to-layer="4537" to-port="3" />
-		<edge from-layer="4537" from-port="4" to-layer="4538" to-port="1" />
-		<edge from-layer="4538" from-port="2" to-layer="4547" to-port="0" />
-		<edge from-layer="4539" from-port="0" to-layer="4540" to-port="0" />
-		<edge from-layer="4540" from-port="1" to-layer="4543" to-port="0" />
-		<edge from-layer="4541" from-port="0" to-layer="4543" to-port="1" />
-		<edge from-layer="4542" from-port="0" to-layer="4543" to-port="3" />
-		<edge from-layer="4543" from-port="4" to-layer="4545" to-port="0" />
-		<edge from-layer="4544" from-port="0" to-layer="4545" to-port="2" />
-		<edge from-layer="4545" from-port="3" to-layer="4546" to-port="0" />
-		<edge from-layer="4546" from-port="2" to-layer="4547" to-port="1" />
-		<edge from-layer="4546" from-port="2" to-layer="4580" to-port="1" />
-		<edge from-layer="4547" from-port="2" to-layer="4548" to-port="1" />
-		<edge from-layer="4548" from-port="2" to-layer="4553" to-port="0" />
-		<edge from-layer="4549" from-port="0" to-layer="4552" to-port="1" />
-		<edge from-layer="4550" from-port="0" to-layer="4552" to-port="2" />
-		<edge from-layer="4551" from-port="0" to-layer="4552" to-port="3" />
-		<edge from-layer="4552" from-port="4" to-layer="4553" to-port="1" />
-		<edge from-layer="4553" from-port="2" to-layer="4611" to-port="0" />
-		<edge from-layer="4553" from-port="2" to-layer="4644" to-port="0" />
-		<edge from-layer="4554" from-port="0" to-layer="4557" to-port="1" />
-		<edge from-layer="4555" from-port="0" to-layer="4557" to-port="2" />
-		<edge from-layer="4556" from-port="0" to-layer="4557" to-port="3" />
-		<edge from-layer="4557" from-port="4" to-layer="4578" to-port="0" />
-		<edge from-layer="4557" from-port="4" to-layer="4572" to-port="0" />
-		<edge from-layer="4557" from-port="4" to-layer="4561" to-port="0" />
-		<edge from-layer="4557" from-port="4" to-layer="4558" to-port="0" />
-		<edge from-layer="4558" from-port="2" to-layer="4581" to-port="0" />
-		<edge from-layer="4559" from-port="0" to-layer="4569" to-port="0" />
-		<edge from-layer="4560" from-port="0" to-layer="4569" to-port="1" />
-		<edge from-layer="4561" from-port="1" to-layer="4563" to-port="0" />
-		<edge from-layer="4562" from-port="0" to-layer="4563" to-port="2" />
-		<edge from-layer="4563" from-port="3" to-layer="4564" to-port="0" />
-		<edge from-layer="4564" from-port="2" to-layer="4565" to-port="0" />
-		<edge from-layer="4565" from-port="1" to-layer="4567" to-port="0" />
-		<edge from-layer="4566" from-port="0" to-layer="4567" to-port="1" />
-		<edge from-layer="4567" from-port="2" to-layer="4569" to-port="2" />
-		<edge from-layer="4568" from-port="0" to-layer="4569" to-port="3" />
-		<edge from-layer="4569" from-port="4" to-layer="4578" to-port="2" />
-		<edge from-layer="4569" from-port="4" to-layer="4572" to-port="1" />
-		<edge from-layer="4570" from-port="0" to-layer="4572" to-port="2" />
-		<edge from-layer="4571" from-port="0" to-layer="4572" to-port="3" />
-		<edge from-layer="4572" from-port="4" to-layer="4575" to-port="0" />
-		<edge from-layer="4573" from-port="0" to-layer="4574" to-port="0" />
-		<edge from-layer="4574" from-port="1" to-layer="4575" to-port="1" />
-		<edge from-layer="4575" from-port="2" to-layer="4579" to-port="0" />
-		<edge from-layer="4576" from-port="0" to-layer="4578" to-port="1" />
-		<edge from-layer="4577" from-port="0" to-layer="4578" to-port="3" />
-		<edge from-layer="4578" from-port="4" to-layer="4579" to-port="1" />
-		<edge from-layer="4579" from-port="2" to-layer="4580" to-port="0" />
-		<edge from-layer="4580" from-port="2" to-layer="4581" to-port="1" />
-		<edge from-layer="4581" from-port="2" to-layer="4586" to-port="0" />
-		<edge from-layer="4582" from-port="0" to-layer="4585" to-port="1" />
-		<edge from-layer="4583" from-port="0" to-layer="4585" to-port="2" />
-		<edge from-layer="4584" from-port="0" to-layer="4585" to-port="3" />
-		<edge from-layer="4585" from-port="4" to-layer="4586" to-port="1" />
-		<edge from-layer="4586" from-port="2" to-layer="4587" to-port="1" />
-		<edge from-layer="4587" from-port="2" to-layer="4608" to-port="0" />
-		<edge from-layer="4587" from-port="2" to-layer="4644" to-port="1" />
-		<edge from-layer="4587" from-port="2" to-layer="7702" to-port="0" />
+		<edge from-layer="4531" from-port="4" to-layer="4532" to-port="1" />
+		<edge from-layer="4532" from-port="2" to-layer="4533" to-port="0" />
+		<edge from-layer="4533" from-port="1" to-layer="4535" to-port="0" />
+		<edge from-layer="4534" from-port="0" to-layer="4535" to-port="2" />
+		<edge from-layer="4535" from-port="3" to-layer="4542" to-port="0" />
+		<edge from-layer="4536" from-port="0" to-layer="4539" to-port="1" />
+		<edge from-layer="4537" from-port="0" to-layer="4539" to-port="2" />
+		<edge from-layer="4538" from-port="0" to-layer="4539" to-port="3" />
+		<edge from-layer="4539" from-port="5" to-layer="4541" to-port="0" />
+		<edge from-layer="4540" from-port="0" to-layer="4541" to-port="1" />
+		<edge from-layer="4541" from-port="2" to-layer="4542" to-port="1" />
+		<edge from-layer="4542" from-port="2" to-layer="4551" to-port="2" />
+		<edge from-layer="4542" from-port="2" to-layer="5118" to-port="0" />
+		<edge from-layer="4543" from-port="0" to-layer="4550" to-port="1" />
+		<edge from-layer="4544" from-port="1" to-layer="4547" to-port="0" />
+		<edge from-layer="4545" from-port="0" to-layer="4547" to-port="1" />
+		<edge from-layer="4546" from-port="0" to-layer="4547" to-port="2" />
+		<edge from-layer="4547" from-port="3" to-layer="4548" to-port="0" />
+		<edge from-layer="4548" from-port="2" to-layer="4550" to-port="2" />
+		<edge from-layer="4549" from-port="0" to-layer="4550" to-port="3" />
+		<edge from-layer="4550" from-port="5" to-layer="4551" to-port="3" />
+		<edge from-layer="4551" from-port="4" to-layer="4553" to-port="0" />
+		<edge from-layer="4552" from-port="0" to-layer="4553" to-port="1" />
+		<edge from-layer="4553" from-port="2" to-layer="4555" to-port="0" />
+		<edge from-layer="4554" from-port="0" to-layer="4555" to-port="1" />
+		<edge from-layer="4555" from-port="2" to-layer="4564" to-port="0" />
+		<edge from-layer="4556" from-port="0" to-layer="4557" to-port="0" />
+		<edge from-layer="4557" from-port="1" to-layer="4560" to-port="0" />
+		<edge from-layer="4558" from-port="0" to-layer="4559" to-port="0" />
+		<edge from-layer="4559" from-port="1" to-layer="4560" to-port="1" />
+		<edge from-layer="4560" from-port="2" to-layer="4562" to-port="0" />
+		<edge from-layer="4561" from-port="0" to-layer="4562" to-port="1" />
+		<edge from-layer="4562" from-port="2" to-layer="4563" to-port="0" />
+		<edge from-layer="4563" from-port="1" to-layer="4564" to-port="1" />
+		<edge from-layer="4564" from-port="2" to-layer="4566" to-port="0" />
+		<edge from-layer="4565" from-port="0" to-layer="4566" to-port="1" />
+		<edge from-layer="4566" from-port="2" to-layer="4567" to-port="1" />
+		<edge from-layer="4567" from-port="2" to-layer="4568" to-port="0" />
+		<edge from-layer="4568" from-port="2" to-layer="4570" to-port="0" />
+		<edge from-layer="4568" from-port="2" to-layer="4726" to-port="1" />
+		<edge from-layer="4569" from-port="0" to-layer="4570" to-port="1" />
+		<edge from-layer="4570" from-port="2" to-layer="4572" to-port="0" />
+		<edge from-layer="4570" from-port="2" to-layer="4601" to-port="0" />
+		<edge from-layer="4571" from-port="0" to-layer="4572" to-port="1" />
+		<edge from-layer="4572" from-port="2" to-layer="4574" to-port="0" />
+		<edge from-layer="4573" from-port="0" to-layer="4574" to-port="1" />
+		<edge from-layer="4574" from-port="2" to-layer="4583" to-port="0" />
+		<edge from-layer="4575" from-port="0" to-layer="4576" to-port="0" />
+		<edge from-layer="4576" from-port="1" to-layer="4579" to-port="0" />
+		<edge from-layer="4577" from-port="0" to-layer="4578" to-port="0" />
+		<edge from-layer="4578" from-port="1" to-layer="4579" to-port="1" />
+		<edge from-layer="4579" from-port="2" to-layer="4581" to-port="0" />
+		<edge from-layer="4580" from-port="0" to-layer="4581" to-port="1" />
+		<edge from-layer="4581" from-port="2" to-layer="4582" to-port="0" />
+		<edge from-layer="4582" from-port="1" to-layer="4583" to-port="1" />
+		<edge from-layer="4583" from-port="2" to-layer="4586" to-port="0" />
+		<edge from-layer="4584" from-port="0" to-layer="4585" to-port="0" />
+		<edge from-layer="4585" from-port="1" to-layer="4586" to-port="1" />
+		<edge from-layer="4586" from-port="2" to-layer="4587" to-port="0" />
+		<edge from-layer="4587" from-port="1" to-layer="4596" to-port="0" />
 		<edge from-layer="4588" from-port="0" to-layer="4589" to-port="0" />
-		<edge from-layer="4589" from-port="1" to-layer="4594" to-port="0" />
-		<edge from-layer="4590" from-port="0" to-layer="4593" to-port="1" />
-		<edge from-layer="4591" from-port="0" to-layer="4593" to-port="2" />
-		<edge from-layer="4592" from-port="0" to-layer="4593" to-port="3" />
-		<edge from-layer="4593" from-port="4" to-layer="4594" to-port="1" />
+		<edge from-layer="4589" from-port="1" to-layer="4592" to-port="0" />
+		<edge from-layer="4590" from-port="0" to-layer="4591" to-port="0" />
+		<edge from-layer="4591" from-port="1" to-layer="4592" to-port="1" />
+		<edge from-layer="4592" from-port="2" to-layer="4594" to-port="0" />
+		<edge from-layer="4593" from-port="0" to-layer="4594" to-port="1" />
 		<edge from-layer="4594" from-port="2" to-layer="4595" to-port="0" />
-		<edge from-layer="4595" from-port="1" to-layer="4597" to-port="0" />
-		<edge from-layer="4596" from-port="0" to-layer="4597" to-port="2" />
-		<edge from-layer="4597" from-port="3" to-layer="4604" to-port="0" />
-		<edge from-layer="4598" from-port="0" to-layer="4601" to-port="1" />
-		<edge from-layer="4599" from-port="0" to-layer="4601" to-port="2" />
-		<edge from-layer="4600" from-port="0" to-layer="4601" to-port="3" />
-		<edge from-layer="4601" from-port="4" to-layer="4603" to-port="0" />
+		<edge from-layer="4595" from-port="1" to-layer="4596" to-port="1" />
+		<edge from-layer="4596" from-port="2" to-layer="4599" to-port="0" />
+		<edge from-layer="4597" from-port="0" to-layer="4598" to-port="0" />
+		<edge from-layer="4598" from-port="1" to-layer="4599" to-port="1" />
+		<edge from-layer="4599" from-port="2" to-layer="4725" to-port="0" />
+		<edge from-layer="4600" from-port="0" to-layer="4601" to-port="1" />
+		<edge from-layer="4601" from-port="2" to-layer="4603" to-port="0" />
 		<edge from-layer="4602" from-port="0" to-layer="4603" to-port="1" />
-		<edge from-layer="4603" from-port="2" to-layer="4604" to-port="1" />
-		<edge from-layer="4604" from-port="2" to-layer="7701" to-port="0" />
-		<edge from-layer="4604" from-port="2" to-layer="4644" to-port="2" />
-		<edge from-layer="4605" from-port="0" to-layer="4623" to-port="0" />
-		<edge from-layer="4606" from-port="0" to-layer="4618" to-port="0" />
-		<edge from-layer="4607" from-port="0" to-layer="4618" to-port="1" />
-		<edge from-layer="4607" from-port="0" to-layer="4621" to-port="1" />
-		<edge from-layer="4608" from-port="1" to-layer="4610" to-port="0" />
-		<edge from-layer="4609" from-port="0" to-layer="4610" to-port="2" />
-		<edge from-layer="4610" from-port="3" to-layer="4614" to-port="0" />
-		<edge from-layer="4610" from-port="3" to-layer="4620" to-port="0" />
-		<edge from-layer="4611" from-port="1" to-layer="4613" to-port="0" />
-		<edge from-layer="4611" from-port="1" to-layer="4639" to-port="0" />
-		<edge from-layer="4612" from-port="0" to-layer="4613" to-port="2" />
-		<edge from-layer="4613" from-port="3" to-layer="4614" to-port="1" />
+		<edge from-layer="4603" from-port="2" to-layer="4612" to-port="0" />
+		<edge from-layer="4604" from-port="0" to-layer="4605" to-port="0" />
+		<edge from-layer="4605" from-port="1" to-layer="4608" to-port="0" />
+		<edge from-layer="4606" from-port="0" to-layer="4607" to-port="0" />
+		<edge from-layer="4607" from-port="1" to-layer="4608" to-port="1" />
+		<edge from-layer="4608" from-port="2" to-layer="4610" to-port="0" />
+		<edge from-layer="4609" from-port="0" to-layer="4610" to-port="1" />
+		<edge from-layer="4610" from-port="2" to-layer="4611" to-port="0" />
+		<edge from-layer="4611" from-port="1" to-layer="4612" to-port="1" />
+		<edge from-layer="4612" from-port="2" to-layer="4614" to-port="0" />
+		<edge from-layer="4613" from-port="0" to-layer="4614" to-port="1" />
 		<edge from-layer="4614" from-port="2" to-layer="4616" to-port="0" />
-		<edge from-layer="4615" from-port="0" to-layer="4620" to-port="1" />
 		<edge from-layer="4615" from-port="0" to-layer="4616" to-port="1" />
-		<edge from-layer="4616" from-port="2" to-layer="4618" to-port="2" />
-		<edge from-layer="4617" from-port="0" to-layer="4618" to-port="3" />
-		<edge from-layer="4617" from-port="0" to-layer="4621" to-port="3" />
-		<edge from-layer="4618" from-port="4" to-layer="4623" to-port="1" />
-		<edge from-layer="4619" from-port="0" to-layer="4621" to-port="0" />
-		<edge from-layer="4620" from-port="2" to-layer="4621" to-port="2" />
-		<edge from-layer="4620" from-port="2" to-layer="4628" to-port="2" />
-		<edge from-layer="4621" from-port="4" to-layer="4623" to-port="2" />
-		<edge from-layer="4622" from-port="0" to-layer="4623" to-port="3" />
-		<edge from-layer="4623" from-port="4" to-layer="4630" to-port="0" />
-		<edge from-layer="4624" from-port="0" to-layer="4630" to-port="1" />
-		<edge from-layer="4625" from-port="0" to-layer="4628" to-port="0" />
-		<edge from-layer="4626" from-port="0" to-layer="4628" to-port="1" />
-		<edge from-layer="4627" from-port="0" to-layer="4628" to-port="3" />
-		<edge from-layer="4628" from-port="4" to-layer="4630" to-port="2" />
-		<edge from-layer="4629" from-port="0" to-layer="4630" to-port="3" />
-		<edge from-layer="4630" from-port="4" to-layer="4634" to-port="0" />
-		<edge from-layer="4631" from-port="0" to-layer="4632" to-port="0" />
-		<edge from-layer="4632" from-port="1" to-layer="4634" to-port="1" />
-		<edge from-layer="4633" from-port="0" to-layer="4634" to-port="2" />
-		<edge from-layer="4634" from-port="3" to-layer="4635" to-port="0" />
-		<edge from-layer="4635" from-port="2" to-layer="4636" to-port="0" />
-		<edge from-layer="4635" from-port="2" to-layer="4643" to-port="0" />
-		<edge from-layer="4636" from-port="1" to-layer="4642" to-port="0" />
-		<edge from-layer="4637" from-port="0" to-layer="4639" to-port="1" />
-		<edge from-layer="4638" from-port="0" to-layer="4639" to-port="2" />
-		<edge from-layer="4639" from-port="3" to-layer="4640" to-port="0" />
-		<edge from-layer="4640" from-port="4" to-layer="4641" to-port="0" />
-		<edge from-layer="4641" from-port="1" to-layer="4642" to-port="1" />
-		<edge from-layer="4642" from-port="2" to-layer="4643" to-port="1" />
-		<edge from-layer="4643" from-port="2" to-layer="4644" to-port="3" />
-		<edge from-layer="4644" from-port="4" to-layer="4646" to-port="0" />
-		<edge from-layer="4645" from-port="0" to-layer="4646" to-port="1" />
-		<edge from-layer="4646" from-port="2" to-layer="4648" to-port="0" />
-		<edge from-layer="4647" from-port="0" to-layer="4648" to-port="1" />
-		<edge from-layer="4648" from-port="2" to-layer="4657" to-port="0" />
-		<edge from-layer="4649" from-port="0" to-layer="4650" to-port="0" />
-		<edge from-layer="4650" from-port="1" to-layer="4653" to-port="0" />
-		<edge from-layer="4651" from-port="0" to-layer="4652" to-port="0" />
-		<edge from-layer="4652" from-port="1" to-layer="4653" to-port="1" />
-		<edge from-layer="4653" from-port="2" to-layer="4655" to-port="0" />
-		<edge from-layer="4654" from-port="0" to-layer="4655" to-port="1" />
-		<edge from-layer="4655" from-port="2" to-layer="4656" to-port="0" />
-		<edge from-layer="4656" from-port="1" to-layer="4657" to-port="1" />
-		<edge from-layer="4657" from-port="2" to-layer="4660" to-port="0" />
-		<edge from-layer="4658" from-port="0" to-layer="4659" to-port="0" />
-		<edge from-layer="4659" from-port="1" to-layer="4660" to-port="1" />
-		<edge from-layer="4660" from-port="2" to-layer="4661" to-port="1" />
-		<edge from-layer="4661" from-port="2" to-layer="4662" to-port="0" />
-		<edge from-layer="4662" from-port="2" to-layer="4904" to-port="1" />
-		<edge from-layer="4662" from-port="2" to-layer="4664" to-port="0" />
-		<edge from-layer="4663" from-port="0" to-layer="4664" to-port="1" />
-		<edge from-layer="4664" from-port="2" to-layer="4698" to-port="0" />
-		<edge from-layer="4664" from-port="2" to-layer="4667" to-port="0" />
-		<edge from-layer="4665" from-port="0" to-layer="4666" to-port="0" />
-		<edge from-layer="4666" from-port="1" to-layer="4667" to-port="1" />
-		<edge from-layer="4667" from-port="2" to-layer="4670" to-port="0" />
-		<edge from-layer="4668" from-port="0" to-layer="4669" to-port="0" />
-		<edge from-layer="4669" from-port="1" to-layer="4670" to-port="1" />
-		<edge from-layer="4670" from-port="2" to-layer="4679" to-port="0" />
-		<edge from-layer="4671" from-port="0" to-layer="4672" to-port="0" />
-		<edge from-layer="4672" from-port="1" to-layer="4675" to-port="0" />
-		<edge from-layer="4673" from-port="0" to-layer="4674" to-port="0" />
-		<edge from-layer="4674" from-port="1" to-layer="4675" to-port="1" />
-		<edge from-layer="4675" from-port="2" to-layer="4677" to-port="0" />
-		<edge from-layer="4676" from-port="0" to-layer="4677" to-port="1" />
-		<edge from-layer="4677" from-port="2" to-layer="4678" to-port="0" />
-		<edge from-layer="4678" from-port="1" to-layer="4679" to-port="1" />
-		<edge from-layer="4679" from-port="2" to-layer="4682" to-port="0" />
-		<edge from-layer="4680" from-port="0" to-layer="4681" to-port="0" />
-		<edge from-layer="4681" from-port="1" to-layer="4682" to-port="1" />
-		<edge from-layer="4682" from-port="2" to-layer="4683" to-port="0" />
-		<edge from-layer="4683" from-port="1" to-layer="4692" to-port="0" />
-		<edge from-layer="4684" from-port="0" to-layer="4685" to-port="0" />
-		<edge from-layer="4685" from-port="1" to-layer="4688" to-port="0" />
-		<edge from-layer="4686" from-port="0" to-layer="4687" to-port="0" />
-		<edge from-layer="4687" from-port="1" to-layer="4688" to-port="1" />
-		<edge from-layer="4688" from-port="2" to-layer="4690" to-port="0" />
-		<edge from-layer="4689" from-port="0" to-layer="4690" to-port="1" />
+		<edge from-layer="4616" from-port="2" to-layer="4697" to-port="0" />
+		<edge from-layer="4616" from-port="2" to-layer="4658" to-port="0" />
+		<edge from-layer="4616" from-port="2" to-layer="4620" to-port="0" />
+		<edge from-layer="4617" from-port="0" to-layer="4620" to-port="1" />
+		<edge from-layer="4618" from-port="0" to-layer="4620" to-port="2" />
+		<edge from-layer="4619" from-port="0" to-layer="4620" to-port="3" />
+		<edge from-layer="4620" from-port="5" to-layer="4622" to-port="0" />
+		<edge from-layer="4621" from-port="0" to-layer="4622" to-port="1" />
+		<edge from-layer="4622" from-port="2" to-layer="4644" to-port="0" />
+		<edge from-layer="4622" from-port="2" to-layer="4626" to-port="0" />
+		<edge from-layer="4623" from-port="0" to-layer="4626" to-port="1" />
+		<edge from-layer="4624" from-port="0" to-layer="4626" to-port="2" />
+		<edge from-layer="4625" from-port="0" to-layer="4626" to-port="3" />
+		<edge from-layer="4626" from-port="5" to-layer="4627" to-port="0" />
+		<edge from-layer="4626" from-port="5" to-layer="4631" to-port="0" />
+		<edge from-layer="4626" from-port="5" to-layer="4637" to-port="0" />
+		<edge from-layer="4627" from-port="2" to-layer="4640" to-port="0" />
+		<edge from-layer="4628" from-port="0" to-layer="4631" to-port="1" />
+		<edge from-layer="4629" from-port="0" to-layer="4631" to-port="2" />
+		<edge from-layer="4630" from-port="0" to-layer="4631" to-port="3" />
+		<edge from-layer="4631" from-port="5" to-layer="4633" to-port="0" />
+		<edge from-layer="4632" from-port="0" to-layer="4633" to-port="1" />
+		<edge from-layer="4633" from-port="2" to-layer="4638" to-port="0" />
+		<edge from-layer="4634" from-port="0" to-layer="4637" to-port="1" />
+		<edge from-layer="4635" from-port="0" to-layer="4637" to-port="2" />
+		<edge from-layer="4636" from-port="0" to-layer="4637" to-port="3" />
+		<edge from-layer="4637" from-port="5" to-layer="4638" to-port="1" />
+		<edge from-layer="4638" from-port="2" to-layer="4639" to-port="0" />
+		<edge from-layer="4639" from-port="2" to-layer="4640" to-port="1" />
+		<edge from-layer="4640" from-port="2" to-layer="4645" to-port="0" />
+		<edge from-layer="4641" from-port="0" to-layer="4644" to-port="1" />
+		<edge from-layer="4642" from-port="0" to-layer="4644" to-port="2" />
+		<edge from-layer="4643" from-port="0" to-layer="4644" to-port="3" />
+		<edge from-layer="4644" from-port="5" to-layer="4645" to-port="1" />
+		<edge from-layer="4645" from-port="2" to-layer="4709" to-port="0" />
+		<edge from-layer="4646" from-port="0" to-layer="4651" to-port="0" />
+		<edge from-layer="4647" from-port="0" to-layer="4650" to-port="1" />
+		<edge from-layer="4648" from-port="0" to-layer="4650" to-port="2" />
+		<edge from-layer="4649" from-port="0" to-layer="4650" to-port="3" />
+		<edge from-layer="4650" from-port="4" to-layer="4651" to-port="1" />
+		<edge from-layer="4651" from-port="2" to-layer="4652" to-port="0" />
+		<edge from-layer="4652" from-port="1" to-layer="4654" to-port="0" />
+		<edge from-layer="4653" from-port="0" to-layer="4654" to-port="2" />
+		<edge from-layer="4654" from-port="3" to-layer="4684" to-port="0" />
+		<edge from-layer="4654" from-port="3" to-layer="4702" to-port="0" />
+		<edge from-layer="4655" from-port="0" to-layer="4658" to-port="1" />
+		<edge from-layer="4656" from-port="0" to-layer="4658" to-port="2" />
+		<edge from-layer="4657" from-port="0" to-layer="4658" to-port="3" />
+		<edge from-layer="4658" from-port="5" to-layer="4660" to-port="0" />
+		<edge from-layer="4659" from-port="0" to-layer="4660" to-port="1" />
+		<edge from-layer="4660" from-port="2" to-layer="4664" to-port="0" />
+		<edge from-layer="4660" from-port="2" to-layer="4682" to-port="0" />
+		<edge from-layer="4661" from-port="0" to-layer="4664" to-port="1" />
+		<edge from-layer="4662" from-port="0" to-layer="4664" to-port="2" />
+		<edge from-layer="4663" from-port="0" to-layer="4664" to-port="3" />
+		<edge from-layer="4664" from-port="5" to-layer="4665" to-port="0" />
+		<edge from-layer="4664" from-port="5" to-layer="4675" to-port="0" />
+		<edge from-layer="4664" from-port="5" to-layer="4669" to-port="0" />
+		<edge from-layer="4665" from-port="2" to-layer="4678" to-port="0" />
+		<edge from-layer="4666" from-port="0" to-layer="4669" to-port="1" />
+		<edge from-layer="4667" from-port="0" to-layer="4669" to-port="2" />
+		<edge from-layer="4668" from-port="0" to-layer="4669" to-port="3" />
+		<edge from-layer="4669" from-port="5" to-layer="4671" to-port="0" />
+		<edge from-layer="4670" from-port="0" to-layer="4671" to-port="1" />
+		<edge from-layer="4671" from-port="2" to-layer="4676" to-port="0" />
+		<edge from-layer="4672" from-port="0" to-layer="4675" to-port="1" />
+		<edge from-layer="4673" from-port="0" to-layer="4675" to-port="2" />
+		<edge from-layer="4674" from-port="0" to-layer="4675" to-port="3" />
+		<edge from-layer="4675" from-port="5" to-layer="4676" to-port="1" />
+		<edge from-layer="4676" from-port="2" to-layer="4677" to-port="0" />
+		<edge from-layer="4677" from-port="2" to-layer="4678" to-port="1" />
+		<edge from-layer="4678" from-port="2" to-layer="4683" to-port="0" />
+		<edge from-layer="4679" from-port="0" to-layer="4682" to-port="1" />
+		<edge from-layer="4680" from-port="0" to-layer="4682" to-port="2" />
+		<edge from-layer="4681" from-port="0" to-layer="4682" to-port="3" />
+		<edge from-layer="4682" from-port="5" to-layer="4683" to-port="1" />
+		<edge from-layer="4683" from-port="2" to-layer="4684" to-port="1" />
+		<edge from-layer="4684" from-port="2" to-layer="5117" to-port="0" />
+		<edge from-layer="4684" from-port="2" to-layer="4709" to-port="1" />
+		<edge from-layer="4685" from-port="0" to-layer="4690" to-port="0" />
+		<edge from-layer="4686" from-port="0" to-layer="4689" to-port="1" />
+		<edge from-layer="4687" from-port="0" to-layer="4689" to-port="2" />
+		<edge from-layer="4688" from-port="0" to-layer="4689" to-port="3" />
+		<edge from-layer="4689" from-port="4" to-layer="4690" to-port="1" />
 		<edge from-layer="4690" from-port="2" to-layer="4691" to-port="0" />
-		<edge from-layer="4691" from-port="1" to-layer="4692" to-port="1" />
-		<edge from-layer="4692" from-port="2" to-layer="4695" to-port="0" />
-		<edge from-layer="4693" from-port="0" to-layer="4694" to-port="0" />
-		<edge from-layer="4694" from-port="1" to-layer="4695" to-port="1" />
-		<edge from-layer="4695" from-port="2" to-layer="4903" to-port="0" />
-		<edge from-layer="4696" from-port="0" to-layer="4697" to-port="0" />
-		<edge from-layer="4697" from-port="1" to-layer="4698" to-port="1" />
-		<edge from-layer="4698" from-port="2" to-layer="4701" to-port="0" />
-		<edge from-layer="4699" from-port="0" to-layer="4700" to-port="0" />
-		<edge from-layer="4700" from-port="1" to-layer="4701" to-port="1" />
-		<edge from-layer="4701" from-port="2" to-layer="4710" to-port="0" />
-		<edge from-layer="4702" from-port="0" to-layer="4703" to-port="0" />
-		<edge from-layer="4703" from-port="1" to-layer="4706" to-port="0" />
-		<edge from-layer="4704" from-port="0" to-layer="4705" to-port="0" />
-		<edge from-layer="4705" from-port="1" to-layer="4706" to-port="1" />
-		<edge from-layer="4706" from-port="2" to-layer="4708" to-port="0" />
-		<edge from-layer="4707" from-port="0" to-layer="4708" to-port="1" />
-		<edge from-layer="4708" from-port="2" to-layer="4709" to-port="0" />
-		<edge from-layer="4709" from-port="1" to-layer="4710" to-port="1" />
-		<edge from-layer="4710" from-port="2" to-layer="4713" to-port="0" />
-		<edge from-layer="4711" from-port="0" to-layer="4712" to-port="0" />
-		<edge from-layer="4712" from-port="1" to-layer="4713" to-port="1" />
-		<edge from-layer="4713" from-port="2" to-layer="4715" to-port="0" />
-		<edge from-layer="4714" from-port="0" to-layer="4715" to-port="1" />
-		<edge from-layer="4715" from-port="2" to-layer="4843" to-port="0" />
-		<edge from-layer="4715" from-port="2" to-layer="4732" to-port="0" />
-		<edge from-layer="4715" from-port="2" to-layer="4719" to-port="0" />
-		<edge from-layer="4716" from-port="0" to-layer="4719" to-port="1" />
-		<edge from-layer="4717" from-port="0" to-layer="4719" to-port="2" />
-		<edge from-layer="4718" from-port="0" to-layer="4719" to-port="3" />
-		<edge from-layer="4719" from-port="4" to-layer="4721" to-port="0" />
-		<edge from-layer="4720" from-port="0" to-layer="4721" to-port="1" />
-		<edge from-layer="4721" from-port="2" to-layer="4794" to-port="0" />
-		<edge from-layer="4721" from-port="2" to-layer="4725" to-port="0" />
-		<edge from-layer="4722" from-port="0" to-layer="4725" to-port="1" />
-		<edge from-layer="4723" from-port="0" to-layer="4725" to-port="2" />
-		<edge from-layer="4724" from-port="0" to-layer="4725" to-port="3" />
-		<edge from-layer="4725" from-port="4" to-layer="4762" to-port="0" />
-		<edge from-layer="4725" from-port="4" to-layer="4759" to-port="0" />
-		<edge from-layer="4725" from-port="4" to-layer="4779" to-port="0" />
-		<edge from-layer="4725" from-port="4" to-layer="4773" to-port="0" />
-		<edge from-layer="4726" from-port="0" to-layer="4727" to-port="0" />
-		<edge from-layer="4727" from-port="1" to-layer="4755" to-port="0" />
-		<edge from-layer="4728" from-port="0" to-layer="4755" to-port="1" />
-		<edge from-layer="4729" from-port="0" to-layer="4732" to-port="1" />
-		<edge from-layer="4730" from-port="0" to-layer="4732" to-port="2" />
-		<edge from-layer="4731" from-port="0" to-layer="4732" to-port="3" />
-		<edge from-layer="4732" from-port="4" to-layer="4734" to-port="0" />
-		<edge from-layer="4733" from-port="0" to-layer="4734" to-port="1" />
-		<edge from-layer="4734" from-port="2" to-layer="4799" to-port="0" />
-		<edge from-layer="4734" from-port="2" to-layer="4827" to-port="0" />
-		<edge from-layer="4734" from-port="2" to-layer="4735" to-port="0" />
-		<edge from-layer="4735" from-port="1" to-layer="4737" to-port="0" />
-		<edge from-layer="4736" from-port="0" to-layer="4737" to-port="2" />
-		<edge from-layer="4737" from-port="3" to-layer="4751" to-port="0" />
-		<edge from-layer="4738" from-port="0" to-layer="4739" to-port="0" />
-		<edge from-layer="4739" from-port="1" to-layer="4744" to-port="0" />
-		<edge from-layer="4740" from-port="0" to-layer="4743" to-port="1" />
-		<edge from-layer="4741" from-port="0" to-layer="4743" to-port="2" />
-		<edge from-layer="4742" from-port="0" to-layer="4743" to-port="3" />
-		<edge from-layer="4743" from-port="4" to-layer="4744" to-port="1" />
+		<edge from-layer="4691" from-port="1" to-layer="4693" to-port="0" />
+		<edge from-layer="4692" from-port="0" to-layer="4693" to-port="2" />
+		<edge from-layer="4693" from-port="3" to-layer="4700" to-port="0" />
+		<edge from-layer="4694" from-port="0" to-layer="4697" to-port="1" />
+		<edge from-layer="4695" from-port="0" to-layer="4697" to-port="2" />
+		<edge from-layer="4696" from-port="0" to-layer="4697" to-port="3" />
+		<edge from-layer="4697" from-port="5" to-layer="4699" to-port="0" />
+		<edge from-layer="4698" from-port="0" to-layer="4699" to-port="1" />
+		<edge from-layer="4699" from-port="2" to-layer="4700" to-port="1" />
+		<edge from-layer="4700" from-port="2" to-layer="4709" to-port="2" />
+		<edge from-layer="4700" from-port="2" to-layer="5116" to-port="0" />
+		<edge from-layer="4701" from-port="0" to-layer="4708" to-port="1" />
+		<edge from-layer="4702" from-port="1" to-layer="4705" to-port="0" />
+		<edge from-layer="4703" from-port="0" to-layer="4705" to-port="1" />
+		<edge from-layer="4704" from-port="0" to-layer="4705" to-port="2" />
+		<edge from-layer="4705" from-port="3" to-layer="4706" to-port="0" />
+		<edge from-layer="4706" from-port="2" to-layer="4708" to-port="2" />
+		<edge from-layer="4707" from-port="0" to-layer="4708" to-port="3" />
+		<edge from-layer="4708" from-port="5" to-layer="4709" to-port="3" />
+		<edge from-layer="4709" from-port="4" to-layer="4711" to-port="0" />
+		<edge from-layer="4710" from-port="0" to-layer="4711" to-port="1" />
+		<edge from-layer="4711" from-port="2" to-layer="4713" to-port="0" />
+		<edge from-layer="4712" from-port="0" to-layer="4713" to-port="1" />
+		<edge from-layer="4713" from-port="2" to-layer="4722" to-port="0" />
+		<edge from-layer="4714" from-port="0" to-layer="4715" to-port="0" />
+		<edge from-layer="4715" from-port="1" to-layer="4718" to-port="0" />
+		<edge from-layer="4716" from-port="0" to-layer="4717" to-port="0" />
+		<edge from-layer="4717" from-port="1" to-layer="4718" to-port="1" />
+		<edge from-layer="4718" from-port="2" to-layer="4720" to-port="0" />
+		<edge from-layer="4719" from-port="0" to-layer="4720" to-port="1" />
+		<edge from-layer="4720" from-port="2" to-layer="4721" to-port="0" />
+		<edge from-layer="4721" from-port="1" to-layer="4722" to-port="1" />
+		<edge from-layer="4722" from-port="2" to-layer="4724" to-port="0" />
+		<edge from-layer="4723" from-port="0" to-layer="4724" to-port="1" />
+		<edge from-layer="4724" from-port="2" to-layer="4725" to-port="1" />
+		<edge from-layer="4725" from-port="2" to-layer="4726" to-port="0" />
+		<edge from-layer="4726" from-port="2" to-layer="4728" to-port="0" />
+		<edge from-layer="4726" from-port="2" to-layer="4884" to-port="1" />
+		<edge from-layer="4727" from-port="0" to-layer="4728" to-port="1" />
+		<edge from-layer="4728" from-port="2" to-layer="4730" to-port="0" />
+		<edge from-layer="4728" from-port="2" to-layer="4759" to-port="0" />
+		<edge from-layer="4729" from-port="0" to-layer="4730" to-port="1" />
+		<edge from-layer="4730" from-port="2" to-layer="4732" to-port="0" />
+		<edge from-layer="4731" from-port="0" to-layer="4732" to-port="1" />
+		<edge from-layer="4732" from-port="2" to-layer="4741" to-port="0" />
+		<edge from-layer="4733" from-port="0" to-layer="4734" to-port="0" />
+		<edge from-layer="4734" from-port="1" to-layer="4737" to-port="0" />
+		<edge from-layer="4735" from-port="0" to-layer="4736" to-port="0" />
+		<edge from-layer="4736" from-port="1" to-layer="4737" to-port="1" />
+		<edge from-layer="4737" from-port="2" to-layer="4739" to-port="0" />
+		<edge from-layer="4738" from-port="0" to-layer="4739" to-port="1" />
+		<edge from-layer="4739" from-port="2" to-layer="4740" to-port="0" />
+		<edge from-layer="4740" from-port="1" to-layer="4741" to-port="1" />
+		<edge from-layer="4741" from-port="2" to-layer="4744" to-port="0" />
+		<edge from-layer="4742" from-port="0" to-layer="4743" to-port="0" />
+		<edge from-layer="4743" from-port="1" to-layer="4744" to-port="1" />
 		<edge from-layer="4744" from-port="2" to-layer="4745" to-port="0" />
-		<edge from-layer="4745" from-port="1" to-layer="4747" to-port="0" />
-		<edge from-layer="4746" from-port="0" to-layer="4747" to-port="2" />
-		<edge from-layer="4747" from-port="3" to-layer="4829" to-port="0" />
-		<edge from-layer="4747" from-port="3" to-layer="4748" to-port="0" />
-		<edge from-layer="4748" from-port="1" to-layer="4750" to-port="0" />
-		<edge from-layer="4749" from-port="0" to-layer="4750" to-port="2" />
-		<edge from-layer="4750" from-port="3" to-layer="4751" to-port="1" />
-		<edge from-layer="4751" from-port="2" to-layer="4753" to-port="0" />
-		<edge from-layer="4752" from-port="0" to-layer="4753" to-port="1" />
-		<edge from-layer="4753" from-port="2" to-layer="4755" to-port="2" />
-		<edge from-layer="4753" from-port="2" to-layer="4785" to-port="2" />
-		<edge from-layer="4754" from-port="0" to-layer="4755" to-port="3" />
-		<edge from-layer="4755" from-port="4" to-layer="4757" to-port="0" />
-		<edge from-layer="4756" from-port="0" to-layer="4757" to-port="2" />
-		<edge from-layer="4757" from-port="3" to-layer="4758" to-port="0" />
-		<edge from-layer="4758" from-port="2" to-layer="4759" to-port="1" />
-		<edge from-layer="4758" from-port="2" to-layer="4800" to-port="1" />
-		<edge from-layer="4759" from-port="2" to-layer="4790" to-port="0" />
-		<edge from-layer="4760" from-port="0" to-layer="4770" to-port="0" />
-		<edge from-layer="4761" from-port="0" to-layer="4770" to-port="1" />
-		<edge from-layer="4762" from-port="1" to-layer="4764" to-port="0" />
-		<edge from-layer="4763" from-port="0" to-layer="4764" to-port="2" />
-		<edge from-layer="4764" from-port="3" to-layer="4765" to-port="0" />
-		<edge from-layer="4765" from-port="2" to-layer="4766" to-port="0" />
-		<edge from-layer="4766" from-port="1" to-layer="4768" to-port="0" />
+		<edge from-layer="4745" from-port="1" to-layer="4754" to-port="0" />
+		<edge from-layer="4746" from-port="0" to-layer="4747" to-port="0" />
+		<edge from-layer="4747" from-port="1" to-layer="4750" to-port="0" />
+		<edge from-layer="4748" from-port="0" to-layer="4749" to-port="0" />
+		<edge from-layer="4749" from-port="1" to-layer="4750" to-port="1" />
+		<edge from-layer="4750" from-port="2" to-layer="4752" to-port="0" />
+		<edge from-layer="4751" from-port="0" to-layer="4752" to-port="1" />
+		<edge from-layer="4752" from-port="2" to-layer="4753" to-port="0" />
+		<edge from-layer="4753" from-port="1" to-layer="4754" to-port="1" />
+		<edge from-layer="4754" from-port="2" to-layer="4757" to-port="0" />
+		<edge from-layer="4755" from-port="0" to-layer="4756" to-port="0" />
+		<edge from-layer="4756" from-port="1" to-layer="4757" to-port="1" />
+		<edge from-layer="4757" from-port="2" to-layer="4883" to-port="0" />
+		<edge from-layer="4758" from-port="0" to-layer="4759" to-port="1" />
+		<edge from-layer="4759" from-port="2" to-layer="4761" to-port="0" />
+		<edge from-layer="4760" from-port="0" to-layer="4761" to-port="1" />
+		<edge from-layer="4761" from-port="2" to-layer="4770" to-port="0" />
+		<edge from-layer="4762" from-port="0" to-layer="4763" to-port="0" />
+		<edge from-layer="4763" from-port="1" to-layer="4766" to-port="0" />
+		<edge from-layer="4764" from-port="0" to-layer="4765" to-port="0" />
+		<edge from-layer="4765" from-port="1" to-layer="4766" to-port="1" />
+		<edge from-layer="4766" from-port="2" to-layer="4768" to-port="0" />
 		<edge from-layer="4767" from-port="0" to-layer="4768" to-port="1" />
-		<edge from-layer="4768" from-port="2" to-layer="4770" to-port="2" />
-		<edge from-layer="4769" from-port="0" to-layer="4770" to-port="3" />
-		<edge from-layer="4770" from-port="4" to-layer="4779" to-port="2" />
-		<edge from-layer="4770" from-port="4" to-layer="4773" to-port="1" />
-		<edge from-layer="4771" from-port="0" to-layer="4773" to-port="2" />
-		<edge from-layer="4772" from-port="0" to-layer="4773" to-port="3" />
-		<edge from-layer="4773" from-port="4" to-layer="4776" to-port="0" />
-		<edge from-layer="4774" from-port="0" to-layer="4775" to-port="0" />
-		<edge from-layer="4775" from-port="1" to-layer="4776" to-port="1" />
-		<edge from-layer="4776" from-port="2" to-layer="4780" to-port="0" />
-		<edge from-layer="4777" from-port="0" to-layer="4779" to-port="1" />
-		<edge from-layer="4778" from-port="0" to-layer="4779" to-port="3" />
-		<edge from-layer="4779" from-port="4" to-layer="4780" to-port="1" />
-		<edge from-layer="4780" from-port="2" to-layer="4789" to-port="0" />
-		<edge from-layer="4781" from-port="0" to-layer="4782" to-port="0" />
-		<edge from-layer="4782" from-port="1" to-layer="4785" to-port="0" />
-		<edge from-layer="4783" from-port="0" to-layer="4785" to-port="1" />
-		<edge from-layer="4784" from-port="0" to-layer="4785" to-port="3" />
-		<edge from-layer="4785" from-port="4" to-layer="4787" to-port="0" />
-		<edge from-layer="4786" from-port="0" to-layer="4787" to-port="2" />
-		<edge from-layer="4787" from-port="3" to-layer="4788" to-port="0" />
-		<edge from-layer="4788" from-port="2" to-layer="4822" to-port="1" />
-		<edge from-layer="4788" from-port="2" to-layer="4789" to-port="1" />
-		<edge from-layer="4789" from-port="2" to-layer="4790" to-port="1" />
-		<edge from-layer="4790" from-port="2" to-layer="4795" to-port="0" />
-		<edge from-layer="4791" from-port="0" to-layer="4794" to-port="1" />
-		<edge from-layer="4792" from-port="0" to-layer="4794" to-port="2" />
-		<edge from-layer="4793" from-port="0" to-layer="4794" to-port="3" />
-		<edge from-layer="4794" from-port="4" to-layer="4795" to-port="1" />
-		<edge from-layer="4795" from-port="2" to-layer="4853" to-port="0" />
-		<edge from-layer="4795" from-port="2" to-layer="4886" to-port="0" />
-		<edge from-layer="4796" from-port="0" to-layer="4799" to-port="1" />
-		<edge from-layer="4797" from-port="0" to-layer="4799" to-port="2" />
-		<edge from-layer="4798" from-port="0" to-layer="4799" to-port="3" />
-		<edge from-layer="4799" from-port="4" to-layer="4803" to-port="0" />
-		<edge from-layer="4799" from-port="4" to-layer="4800" to-port="0" />
-		<edge from-layer="4799" from-port="4" to-layer="4820" to-port="0" />
-		<edge from-layer="4799" from-port="4" to-layer="4814" to-port="0" />
-		<edge from-layer="4800" from-port="2" to-layer="4823" to-port="0" />
-		<edge from-layer="4801" from-port="0" to-layer="4811" to-port="0" />
-		<edge from-layer="4802" from-port="0" to-layer="4811" to-port="1" />
-		<edge from-layer="4803" from-port="1" to-layer="4805" to-port="0" />
-		<edge from-layer="4804" from-port="0" to-layer="4805" to-port="2" />
-		<edge from-layer="4805" from-port="3" to-layer="4806" to-port="0" />
-		<edge from-layer="4806" from-port="2" to-layer="4807" to-port="0" />
-		<edge from-layer="4807" from-port="1" to-layer="4809" to-port="0" />
-		<edge from-layer="4808" from-port="0" to-layer="4809" to-port="1" />
-		<edge from-layer="4809" from-port="2" to-layer="4811" to-port="2" />
-		<edge from-layer="4810" from-port="0" to-layer="4811" to-port="3" />
-		<edge from-layer="4811" from-port="4" to-layer="4814" to-port="1" />
-		<edge from-layer="4811" from-port="4" to-layer="4820" to-port="2" />
-		<edge from-layer="4812" from-port="0" to-layer="4814" to-port="2" />
-		<edge from-layer="4813" from-port="0" to-layer="4814" to-port="3" />
-		<edge from-layer="4814" from-port="4" to-layer="4817" to-port="0" />
-		<edge from-layer="4815" from-port="0" to-layer="4816" to-port="0" />
-		<edge from-layer="4816" from-port="1" to-layer="4817" to-port="1" />
-		<edge from-layer="4817" from-port="2" to-layer="4821" to-port="0" />
-		<edge from-layer="4818" from-port="0" to-layer="4820" to-port="1" />
-		<edge from-layer="4819" from-port="0" to-layer="4820" to-port="3" />
-		<edge from-layer="4820" from-port="4" to-layer="4821" to-port="1" />
-		<edge from-layer="4821" from-port="2" to-layer="4822" to-port="0" />
-		<edge from-layer="4822" from-port="2" to-layer="4823" to-port="1" />
-		<edge from-layer="4823" from-port="2" to-layer="4828" to-port="0" />
+		<edge from-layer="4768" from-port="2" to-layer="4769" to-port="0" />
+		<edge from-layer="4769" from-port="1" to-layer="4770" to-port="1" />
+		<edge from-layer="4770" from-port="2" to-layer="4772" to-port="0" />
+		<edge from-layer="4771" from-port="0" to-layer="4772" to-port="1" />
+		<edge from-layer="4772" from-port="2" to-layer="4774" to-port="0" />
+		<edge from-layer="4773" from-port="0" to-layer="4774" to-port="1" />
+		<edge from-layer="4774" from-port="2" to-layer="4778" to-port="0" />
+		<edge from-layer="4774" from-port="2" to-layer="4816" to-port="0" />
+		<edge from-layer="4774" from-port="2" to-layer="4855" to-port="0" />
+		<edge from-layer="4775" from-port="0" to-layer="4778" to-port="1" />
+		<edge from-layer="4776" from-port="0" to-layer="4778" to-port="2" />
+		<edge from-layer="4777" from-port="0" to-layer="4778" to-port="3" />
+		<edge from-layer="4778" from-port="5" to-layer="4780" to-port="0" />
+		<edge from-layer="4779" from-port="0" to-layer="4780" to-port="1" />
+		<edge from-layer="4780" from-port="2" to-layer="4784" to-port="0" />
+		<edge from-layer="4780" from-port="2" to-layer="4802" to-port="0" />
+		<edge from-layer="4781" from-port="0" to-layer="4784" to-port="1" />
+		<edge from-layer="4782" from-port="0" to-layer="4784" to-port="2" />
+		<edge from-layer="4783" from-port="0" to-layer="4784" to-port="3" />
+		<edge from-layer="4784" from-port="5" to-layer="4795" to-port="0" />
+		<edge from-layer="4784" from-port="5" to-layer="4785" to-port="0" />
+		<edge from-layer="4784" from-port="5" to-layer="4789" to-port="0" />
+		<edge from-layer="4785" from-port="2" to-layer="4798" to-port="0" />
+		<edge from-layer="4786" from-port="0" to-layer="4789" to-port="1" />
+		<edge from-layer="4787" from-port="0" to-layer="4789" to-port="2" />
+		<edge from-layer="4788" from-port="0" to-layer="4789" to-port="3" />
+		<edge from-layer="4789" from-port="5" to-layer="4791" to-port="0" />
+		<edge from-layer="4790" from-port="0" to-layer="4791" to-port="1" />
+		<edge from-layer="4791" from-port="2" to-layer="4796" to-port="0" />
+		<edge from-layer="4792" from-port="0" to-layer="4795" to-port="1" />
+		<edge from-layer="4793" from-port="0" to-layer="4795" to-port="2" />
+		<edge from-layer="4794" from-port="0" to-layer="4795" to-port="3" />
+		<edge from-layer="4795" from-port="5" to-layer="4796" to-port="1" />
+		<edge from-layer="4796" from-port="2" to-layer="4797" to-port="0" />
+		<edge from-layer="4797" from-port="2" to-layer="4798" to-port="1" />
+		<edge from-layer="4798" from-port="2" to-layer="4803" to-port="0" />
+		<edge from-layer="4799" from-port="0" to-layer="4802" to-port="1" />
+		<edge from-layer="4800" from-port="0" to-layer="4802" to-port="2" />
+		<edge from-layer="4801" from-port="0" to-layer="4802" to-port="3" />
+		<edge from-layer="4802" from-port="5" to-layer="4803" to-port="1" />
+		<edge from-layer="4803" from-port="2" to-layer="4867" to-port="0" />
+		<edge from-layer="4804" from-port="0" to-layer="4809" to-port="0" />
+		<edge from-layer="4805" from-port="0" to-layer="4808" to-port="1" />
+		<edge from-layer="4806" from-port="0" to-layer="4808" to-port="2" />
+		<edge from-layer="4807" from-port="0" to-layer="4808" to-port="3" />
+		<edge from-layer="4808" from-port="4" to-layer="4809" to-port="1" />
+		<edge from-layer="4809" from-port="2" to-layer="4810" to-port="0" />
+		<edge from-layer="4810" from-port="1" to-layer="4812" to-port="0" />
+		<edge from-layer="4811" from-port="0" to-layer="4812" to-port="2" />
+		<edge from-layer="4812" from-port="3" to-layer="4842" to-port="0" />
+		<edge from-layer="4812" from-port="3" to-layer="4860" to-port="0" />
+		<edge from-layer="4813" from-port="0" to-layer="4816" to-port="1" />
+		<edge from-layer="4814" from-port="0" to-layer="4816" to-port="2" />
+		<edge from-layer="4815" from-port="0" to-layer="4816" to-port="3" />
+		<edge from-layer="4816" from-port="5" to-layer="4818" to-port="0" />
+		<edge from-layer="4817" from-port="0" to-layer="4818" to-port="1" />
+		<edge from-layer="4818" from-port="2" to-layer="4840" to-port="0" />
+		<edge from-layer="4818" from-port="2" to-layer="4822" to-port="0" />
+		<edge from-layer="4819" from-port="0" to-layer="4822" to-port="1" />
+		<edge from-layer="4820" from-port="0" to-layer="4822" to-port="2" />
+		<edge from-layer="4821" from-port="0" to-layer="4822" to-port="3" />
+		<edge from-layer="4822" from-port="5" to-layer="4833" to-port="0" />
+		<edge from-layer="4822" from-port="5" to-layer="4823" to-port="0" />
+		<edge from-layer="4822" from-port="5" to-layer="4827" to-port="0" />
+		<edge from-layer="4823" from-port="2" to-layer="4836" to-port="0" />
 		<edge from-layer="4824" from-port="0" to-layer="4827" to-port="1" />
 		<edge from-layer="4825" from-port="0" to-layer="4827" to-port="2" />
 		<edge from-layer="4826" from-port="0" to-layer="4827" to-port="3" />
-		<edge from-layer="4827" from-port="4" to-layer="4828" to-port="1" />
-		<edge from-layer="4828" from-port="2" to-layer="4829" to-port="1" />
-		<edge from-layer="4829" from-port="2" to-layer="4886" to-port="1" />
-		<edge from-layer="4829" from-port="2" to-layer="4850" to-port="0" />
-		<edge from-layer="4829" from-port="2" to-layer="7700" to-port="0" />
-		<edge from-layer="4830" from-port="0" to-layer="4831" to-port="0" />
-		<edge from-layer="4831" from-port="1" to-layer="4836" to-port="0" />
-		<edge from-layer="4832" from-port="0" to-layer="4835" to-port="1" />
-		<edge from-layer="4833" from-port="0" to-layer="4835" to-port="2" />
-		<edge from-layer="4834" from-port="0" to-layer="4835" to-port="3" />
-		<edge from-layer="4835" from-port="4" to-layer="4836" to-port="1" />
-		<edge from-layer="4836" from-port="2" to-layer="4837" to-port="0" />
-		<edge from-layer="4837" from-port="1" to-layer="4839" to-port="0" />
-		<edge from-layer="4838" from-port="0" to-layer="4839" to-port="2" />
-		<edge from-layer="4839" from-port="3" to-layer="4846" to-port="0" />
-		<edge from-layer="4840" from-port="0" to-layer="4843" to-port="1" />
-		<edge from-layer="4841" from-port="0" to-layer="4843" to-port="2" />
-		<edge from-layer="4842" from-port="0" to-layer="4843" to-port="3" />
-		<edge from-layer="4843" from-port="4" to-layer="4845" to-port="0" />
-		<edge from-layer="4844" from-port="0" to-layer="4845" to-port="1" />
-		<edge from-layer="4845" from-port="2" to-layer="4846" to-port="1" />
-		<edge from-layer="4846" from-port="2" to-layer="4886" to-port="2" />
-		<edge from-layer="4846" from-port="2" to-layer="7699" to-port="0" />
-		<edge from-layer="4847" from-port="0" to-layer="4865" to-port="0" />
-		<edge from-layer="4848" from-port="0" to-layer="4860" to-port="0" />
-		<edge from-layer="4849" from-port="0" to-layer="4860" to-port="1" />
-		<edge from-layer="4849" from-port="0" to-layer="4863" to-port="1" />
-		<edge from-layer="4850" from-port="1" to-layer="4852" to-port="0" />
-		<edge from-layer="4851" from-port="0" to-layer="4852" to-port="2" />
-		<edge from-layer="4852" from-port="3" to-layer="4856" to-port="0" />
-		<edge from-layer="4852" from-port="3" to-layer="4862" to-port="0" />
-		<edge from-layer="4853" from-port="1" to-layer="4881" to-port="0" />
-		<edge from-layer="4853" from-port="1" to-layer="4855" to-port="0" />
-		<edge from-layer="4854" from-port="0" to-layer="4855" to-port="2" />
-		<edge from-layer="4855" from-port="3" to-layer="4856" to-port="1" />
-		<edge from-layer="4856" from-port="2" to-layer="4858" to-port="0" />
-		<edge from-layer="4857" from-port="0" to-layer="4862" to-port="1" />
-		<edge from-layer="4857" from-port="0" to-layer="4858" to-port="1" />
-		<edge from-layer="4858" from-port="2" to-layer="4860" to-port="2" />
-		<edge from-layer="4859" from-port="0" to-layer="4863" to-port="3" />
-		<edge from-layer="4859" from-port="0" to-layer="4860" to-port="3" />
-		<edge from-layer="4860" from-port="4" to-layer="4865" to-port="1" />
-		<edge from-layer="4861" from-port="0" to-layer="4863" to-port="0" />
-		<edge from-layer="4862" from-port="2" to-layer="4870" to-port="2" />
-		<edge from-layer="4862" from-port="2" to-layer="4863" to-port="2" />
-		<edge from-layer="4863" from-port="4" to-layer="4865" to-port="2" />
-		<edge from-layer="4864" from-port="0" to-layer="4865" to-port="3" />
-		<edge from-layer="4865" from-port="4" to-layer="4872" to-port="0" />
-		<edge from-layer="4866" from-port="0" to-layer="4872" to-port="1" />
-		<edge from-layer="4867" from-port="0" to-layer="4870" to-port="0" />
-		<edge from-layer="4868" from-port="0" to-layer="4870" to-port="1" />
-		<edge from-layer="4869" from-port="0" to-layer="4870" to-port="3" />
-		<edge from-layer="4870" from-port="4" to-layer="4872" to-port="2" />
-		<edge from-layer="4871" from-port="0" to-layer="4872" to-port="3" />
-		<edge from-layer="4872" from-port="4" to-layer="4876" to-port="0" />
-		<edge from-layer="4873" from-port="0" to-layer="4874" to-port="0" />
-		<edge from-layer="4874" from-port="1" to-layer="4876" to-port="1" />
-		<edge from-layer="4875" from-port="0" to-layer="4876" to-port="2" />
-		<edge from-layer="4876" from-port="3" to-layer="4877" to-port="0" />
-		<edge from-layer="4877" from-port="2" to-layer="4885" to-port="0" />
-		<edge from-layer="4877" from-port="2" to-layer="4878" to-port="0" />
-		<edge from-layer="4878" from-port="1" to-layer="4884" to-port="0" />
-		<edge from-layer="4879" from-port="0" to-layer="4881" to-port="1" />
-		<edge from-layer="4880" from-port="0" to-layer="4881" to-port="2" />
-		<edge from-layer="4881" from-port="3" to-layer="4882" to-port="0" />
-		<edge from-layer="4882" from-port="4" to-layer="4883" to-port="0" />
-		<edge from-layer="4883" from-port="1" to-layer="4884" to-port="1" />
-		<edge from-layer="4884" from-port="2" to-layer="4885" to-port="1" />
-		<edge from-layer="4885" from-port="2" to-layer="4886" to-port="3" />
-		<edge from-layer="4886" from-port="4" to-layer="4888" to-port="0" />
+		<edge from-layer="4827" from-port="5" to-layer="4829" to-port="0" />
+		<edge from-layer="4828" from-port="0" to-layer="4829" to-port="1" />
+		<edge from-layer="4829" from-port="2" to-layer="4834" to-port="0" />
+		<edge from-layer="4830" from-port="0" to-layer="4833" to-port="1" />
+		<edge from-layer="4831" from-port="0" to-layer="4833" to-port="2" />
+		<edge from-layer="4832" from-port="0" to-layer="4833" to-port="3" />
+		<edge from-layer="4833" from-port="5" to-layer="4834" to-port="1" />
+		<edge from-layer="4834" from-port="2" to-layer="4835" to-port="0" />
+		<edge from-layer="4835" from-port="2" to-layer="4836" to-port="1" />
+		<edge from-layer="4836" from-port="2" to-layer="4841" to-port="0" />
+		<edge from-layer="4837" from-port="0" to-layer="4840" to-port="1" />
+		<edge from-layer="4838" from-port="0" to-layer="4840" to-port="2" />
+		<edge from-layer="4839" from-port="0" to-layer="4840" to-port="3" />
+		<edge from-layer="4840" from-port="5" to-layer="4841" to-port="1" />
+		<edge from-layer="4841" from-port="2" to-layer="4842" to-port="1" />
+		<edge from-layer="4842" from-port="2" to-layer="4867" to-port="1" />
+		<edge from-layer="4842" from-port="2" to-layer="5115" to-port="0" />
+		<edge from-layer="4843" from-port="0" to-layer="4848" to-port="0" />
+		<edge from-layer="4844" from-port="0" to-layer="4847" to-port="1" />
+		<edge from-layer="4845" from-port="0" to-layer="4847" to-port="2" />
+		<edge from-layer="4846" from-port="0" to-layer="4847" to-port="3" />
+		<edge from-layer="4847" from-port="4" to-layer="4848" to-port="1" />
+		<edge from-layer="4848" from-port="2" to-layer="4849" to-port="0" />
+		<edge from-layer="4849" from-port="1" to-layer="4851" to-port="0" />
+		<edge from-layer="4850" from-port="0" to-layer="4851" to-port="2" />
+		<edge from-layer="4851" from-port="3" to-layer="4858" to-port="0" />
+		<edge from-layer="4852" from-port="0" to-layer="4855" to-port="1" />
+		<edge from-layer="4853" from-port="0" to-layer="4855" to-port="2" />
+		<edge from-layer="4854" from-port="0" to-layer="4855" to-port="3" />
+		<edge from-layer="4855" from-port="5" to-layer="4857" to-port="0" />
+		<edge from-layer="4856" from-port="0" to-layer="4857" to-port="1" />
+		<edge from-layer="4857" from-port="2" to-layer="4858" to-port="1" />
+		<edge from-layer="4858" from-port="2" to-layer="4867" to-port="2" />
+		<edge from-layer="4858" from-port="2" to-layer="5114" to-port="0" />
+		<edge from-layer="4859" from-port="0" to-layer="4866" to-port="1" />
+		<edge from-layer="4860" from-port="1" to-layer="4863" to-port="0" />
+		<edge from-layer="4861" from-port="0" to-layer="4863" to-port="1" />
+		<edge from-layer="4862" from-port="0" to-layer="4863" to-port="2" />
+		<edge from-layer="4863" from-port="3" to-layer="4864" to-port="0" />
+		<edge from-layer="4864" from-port="2" to-layer="4866" to-port="2" />
+		<edge from-layer="4865" from-port="0" to-layer="4866" to-port="3" />
+		<edge from-layer="4866" from-port="5" to-layer="4867" to-port="3" />
+		<edge from-layer="4867" from-port="4" to-layer="4869" to-port="0" />
+		<edge from-layer="4868" from-port="0" to-layer="4869" to-port="1" />
+		<edge from-layer="4869" from-port="2" to-layer="4871" to-port="0" />
+		<edge from-layer="4870" from-port="0" to-layer="4871" to-port="1" />
+		<edge from-layer="4871" from-port="2" to-layer="4880" to-port="0" />
+		<edge from-layer="4872" from-port="0" to-layer="4873" to-port="0" />
+		<edge from-layer="4873" from-port="1" to-layer="4876" to-port="0" />
+		<edge from-layer="4874" from-port="0" to-layer="4875" to-port="0" />
+		<edge from-layer="4875" from-port="1" to-layer="4876" to-port="1" />
+		<edge from-layer="4876" from-port="2" to-layer="4878" to-port="0" />
+		<edge from-layer="4877" from-port="0" to-layer="4878" to-port="1" />
+		<edge from-layer="4878" from-port="2" to-layer="4879" to-port="0" />
+		<edge from-layer="4879" from-port="1" to-layer="4880" to-port="1" />
+		<edge from-layer="4880" from-port="2" to-layer="4882" to-port="0" />
+		<edge from-layer="4881" from-port="0" to-layer="4882" to-port="1" />
+		<edge from-layer="4882" from-port="2" to-layer="4883" to-port="1" />
+		<edge from-layer="4883" from-port="2" to-layer="4884" to-port="0" />
+		<edge from-layer="4884" from-port="2" to-layer="5042" to-port="1" />
+		<edge from-layer="4884" from-port="2" to-layer="4886" to-port="0" />
+		<edge from-layer="4885" from-port="0" to-layer="4886" to-port="1" />
+		<edge from-layer="4886" from-port="2" to-layer="4917" to-port="0" />
+		<edge from-layer="4886" from-port="2" to-layer="4888" to-port="0" />
 		<edge from-layer="4887" from-port="0" to-layer="4888" to-port="1" />
 		<edge from-layer="4888" from-port="2" to-layer="4890" to-port="0" />
 		<edge from-layer="4889" from-port="0" to-layer="4890" to-port="1" />
@@ -127028,3283 +86490,338 @@
 		<edge from-layer="4899" from-port="2" to-layer="4902" to-port="0" />
 		<edge from-layer="4900" from-port="0" to-layer="4901" to-port="0" />
 		<edge from-layer="4901" from-port="1" to-layer="4902" to-port="1" />
-		<edge from-layer="4902" from-port="2" to-layer="4903" to-port="1" />
-		<edge from-layer="4903" from-port="2" to-layer="4904" to-port="0" />
-		<edge from-layer="4904" from-port="2" to-layer="5146" to-port="1" />
-		<edge from-layer="4904" from-port="2" to-layer="4906" to-port="0" />
-		<edge from-layer="4905" from-port="0" to-layer="4906" to-port="1" />
-		<edge from-layer="4906" from-port="2" to-layer="4909" to-port="0" />
-		<edge from-layer="4906" from-port="2" to-layer="4940" to-port="0" />
-		<edge from-layer="4907" from-port="0" to-layer="4908" to-port="0" />
-		<edge from-layer="4908" from-port="1" to-layer="4909" to-port="1" />
-		<edge from-layer="4909" from-port="2" to-layer="4912" to-port="0" />
-		<edge from-layer="4910" from-port="0" to-layer="4911" to-port="0" />
+		<edge from-layer="4902" from-port="2" to-layer="4903" to-port="0" />
+		<edge from-layer="4903" from-port="1" to-layer="4912" to-port="0" />
+		<edge from-layer="4904" from-port="0" to-layer="4905" to-port="0" />
+		<edge from-layer="4905" from-port="1" to-layer="4908" to-port="0" />
+		<edge from-layer="4906" from-port="0" to-layer="4907" to-port="0" />
+		<edge from-layer="4907" from-port="1" to-layer="4908" to-port="1" />
+		<edge from-layer="4908" from-port="2" to-layer="4910" to-port="0" />
+		<edge from-layer="4909" from-port="0" to-layer="4910" to-port="1" />
+		<edge from-layer="4910" from-port="2" to-layer="4911" to-port="0" />
 		<edge from-layer="4911" from-port="1" to-layer="4912" to-port="1" />
-		<edge from-layer="4912" from-port="2" to-layer="4921" to-port="0" />
+		<edge from-layer="4912" from-port="2" to-layer="4915" to-port="0" />
 		<edge from-layer="4913" from-port="0" to-layer="4914" to-port="0" />
-		<edge from-layer="4914" from-port="1" to-layer="4917" to-port="0" />
-		<edge from-layer="4915" from-port="0" to-layer="4916" to-port="0" />
-		<edge from-layer="4916" from-port="1" to-layer="4917" to-port="1" />
+		<edge from-layer="4914" from-port="1" to-layer="4915" to-port="1" />
+		<edge from-layer="4915" from-port="2" to-layer="5041" to-port="0" />
+		<edge from-layer="4916" from-port="0" to-layer="4917" to-port="1" />
 		<edge from-layer="4917" from-port="2" to-layer="4919" to-port="0" />
 		<edge from-layer="4918" from-port="0" to-layer="4919" to-port="1" />
-		<edge from-layer="4919" from-port="2" to-layer="4920" to-port="0" />
-		<edge from-layer="4920" from-port="1" to-layer="4921" to-port="1" />
-		<edge from-layer="4921" from-port="2" to-layer="4924" to-port="0" />
+		<edge from-layer="4919" from-port="2" to-layer="4928" to-port="0" />
+		<edge from-layer="4920" from-port="0" to-layer="4921" to-port="0" />
+		<edge from-layer="4921" from-port="1" to-layer="4924" to-port="0" />
 		<edge from-layer="4922" from-port="0" to-layer="4923" to-port="0" />
 		<edge from-layer="4923" from-port="1" to-layer="4924" to-port="1" />
-		<edge from-layer="4924" from-port="2" to-layer="4925" to-port="0" />
-		<edge from-layer="4925" from-port="1" to-layer="4934" to-port="0" />
-		<edge from-layer="4926" from-port="0" to-layer="4927" to-port="0" />
-		<edge from-layer="4927" from-port="1" to-layer="4930" to-port="0" />
-		<edge from-layer="4928" from-port="0" to-layer="4929" to-port="0" />
-		<edge from-layer="4929" from-port="1" to-layer="4930" to-port="1" />
+		<edge from-layer="4924" from-port="2" to-layer="4926" to-port="0" />
+		<edge from-layer="4925" from-port="0" to-layer="4926" to-port="1" />
+		<edge from-layer="4926" from-port="2" to-layer="4927" to-port="0" />
+		<edge from-layer="4927" from-port="1" to-layer="4928" to-port="1" />
+		<edge from-layer="4928" from-port="2" to-layer="4930" to-port="0" />
+		<edge from-layer="4929" from-port="0" to-layer="4930" to-port="1" />
 		<edge from-layer="4930" from-port="2" to-layer="4932" to-port="0" />
 		<edge from-layer="4931" from-port="0" to-layer="4932" to-port="1" />
-		<edge from-layer="4932" from-port="2" to-layer="4933" to-port="0" />
-		<edge from-layer="4933" from-port="1" to-layer="4934" to-port="1" />
-		<edge from-layer="4934" from-port="2" to-layer="4937" to-port="0" />
-		<edge from-layer="4935" from-port="0" to-layer="4936" to-port="0" />
-		<edge from-layer="4936" from-port="1" to-layer="4937" to-port="1" />
-		<edge from-layer="4937" from-port="2" to-layer="5145" to-port="0" />
-		<edge from-layer="4938" from-port="0" to-layer="4939" to-port="0" />
-		<edge from-layer="4939" from-port="1" to-layer="4940" to-port="1" />
-		<edge from-layer="4940" from-port="2" to-layer="4943" to-port="0" />
-		<edge from-layer="4941" from-port="0" to-layer="4942" to-port="0" />
-		<edge from-layer="4942" from-port="1" to-layer="4943" to-port="1" />
-		<edge from-layer="4943" from-port="2" to-layer="4952" to-port="0" />
-		<edge from-layer="4944" from-port="0" to-layer="4945" to-port="0" />
-		<edge from-layer="4945" from-port="1" to-layer="4948" to-port="0" />
-		<edge from-layer="4946" from-port="0" to-layer="4947" to-port="0" />
-		<edge from-layer="4947" from-port="1" to-layer="4948" to-port="1" />
-		<edge from-layer="4948" from-port="2" to-layer="4950" to-port="0" />
-		<edge from-layer="4949" from-port="0" to-layer="4950" to-port="1" />
-		<edge from-layer="4950" from-port="2" to-layer="4951" to-port="0" />
-		<edge from-layer="4951" from-port="1" to-layer="4952" to-port="1" />
-		<edge from-layer="4952" from-port="2" to-layer="4955" to-port="0" />
-		<edge from-layer="4953" from-port="0" to-layer="4954" to-port="0" />
-		<edge from-layer="4954" from-port="1" to-layer="4955" to-port="1" />
-		<edge from-layer="4955" from-port="2" to-layer="4957" to-port="0" />
-		<edge from-layer="4956" from-port="0" to-layer="4957" to-port="1" />
-		<edge from-layer="4957" from-port="2" to-layer="5085" to-port="0" />
-		<edge from-layer="4957" from-port="2" to-layer="4974" to-port="0" />
-		<edge from-layer="4957" from-port="2" to-layer="4961" to-port="0" />
-		<edge from-layer="4958" from-port="0" to-layer="4961" to-port="1" />
-		<edge from-layer="4959" from-port="0" to-layer="4961" to-port="2" />
-		<edge from-layer="4960" from-port="0" to-layer="4961" to-port="3" />
-		<edge from-layer="4961" from-port="4" to-layer="4963" to-port="0" />
-		<edge from-layer="4962" from-port="0" to-layer="4963" to-port="1" />
-		<edge from-layer="4963" from-port="2" to-layer="4967" to-port="0" />
-		<edge from-layer="4963" from-port="2" to-layer="5036" to-port="0" />
-		<edge from-layer="4964" from-port="0" to-layer="4967" to-port="1" />
-		<edge from-layer="4965" from-port="0" to-layer="4967" to-port="2" />
-		<edge from-layer="4966" from-port="0" to-layer="4967" to-port="3" />
-		<edge from-layer="4967" from-port="4" to-layer="5021" to-port="0" />
-		<edge from-layer="4967" from-port="4" to-layer="5015" to-port="0" />
-		<edge from-layer="4967" from-port="4" to-layer="5004" to-port="0" />
-		<edge from-layer="4967" from-port="4" to-layer="5001" to-port="0" />
-		<edge from-layer="4968" from-port="0" to-layer="4969" to-port="0" />
-		<edge from-layer="4969" from-port="1" to-layer="4997" to-port="0" />
-		<edge from-layer="4970" from-port="0" to-layer="4997" to-port="1" />
+		<edge from-layer="4932" from-port="2" to-layer="4974" to-port="0" />
+		<edge from-layer="4932" from-port="2" to-layer="4936" to-port="0" />
+		<edge from-layer="4932" from-port="2" to-layer="5013" to-port="0" />
+		<edge from-layer="4933" from-port="0" to-layer="4936" to-port="1" />
+		<edge from-layer="4934" from-port="0" to-layer="4936" to-port="2" />
+		<edge from-layer="4935" from-port="0" to-layer="4936" to-port="3" />
+		<edge from-layer="4936" from-port="5" to-layer="4938" to-port="0" />
+		<edge from-layer="4937" from-port="0" to-layer="4938" to-port="1" />
+		<edge from-layer="4938" from-port="2" to-layer="4960" to-port="0" />
+		<edge from-layer="4938" from-port="2" to-layer="4942" to-port="0" />
+		<edge from-layer="4939" from-port="0" to-layer="4942" to-port="1" />
+		<edge from-layer="4940" from-port="0" to-layer="4942" to-port="2" />
+		<edge from-layer="4941" from-port="0" to-layer="4942" to-port="3" />
+		<edge from-layer="4942" from-port="5" to-layer="4947" to-port="0" />
+		<edge from-layer="4942" from-port="5" to-layer="4953" to-port="0" />
+		<edge from-layer="4942" from-port="5" to-layer="4943" to-port="0" />
+		<edge from-layer="4943" from-port="2" to-layer="4956" to-port="0" />
+		<edge from-layer="4944" from-port="0" to-layer="4947" to-port="1" />
+		<edge from-layer="4945" from-port="0" to-layer="4947" to-port="2" />
+		<edge from-layer="4946" from-port="0" to-layer="4947" to-port="3" />
+		<edge from-layer="4947" from-port="5" to-layer="4949" to-port="0" />
+		<edge from-layer="4948" from-port="0" to-layer="4949" to-port="1" />
+		<edge from-layer="4949" from-port="2" to-layer="4954" to-port="0" />
+		<edge from-layer="4950" from-port="0" to-layer="4953" to-port="1" />
+		<edge from-layer="4951" from-port="0" to-layer="4953" to-port="2" />
+		<edge from-layer="4952" from-port="0" to-layer="4953" to-port="3" />
+		<edge from-layer="4953" from-port="5" to-layer="4954" to-port="1" />
+		<edge from-layer="4954" from-port="2" to-layer="4955" to-port="0" />
+		<edge from-layer="4955" from-port="2" to-layer="4956" to-port="1" />
+		<edge from-layer="4956" from-port="2" to-layer="4961" to-port="0" />
+		<edge from-layer="4957" from-port="0" to-layer="4960" to-port="1" />
+		<edge from-layer="4958" from-port="0" to-layer="4960" to-port="2" />
+		<edge from-layer="4959" from-port="0" to-layer="4960" to-port="3" />
+		<edge from-layer="4960" from-port="5" to-layer="4961" to-port="1" />
+		<edge from-layer="4961" from-port="2" to-layer="5025" to-port="0" />
+		<edge from-layer="4962" from-port="0" to-layer="4967" to-port="0" />
+		<edge from-layer="4963" from-port="0" to-layer="4966" to-port="1" />
+		<edge from-layer="4964" from-port="0" to-layer="4966" to-port="2" />
+		<edge from-layer="4965" from-port="0" to-layer="4966" to-port="3" />
+		<edge from-layer="4966" from-port="4" to-layer="4967" to-port="1" />
+		<edge from-layer="4967" from-port="2" to-layer="4968" to-port="0" />
+		<edge from-layer="4968" from-port="1" to-layer="4970" to-port="0" />
+		<edge from-layer="4969" from-port="0" to-layer="4970" to-port="2" />
+		<edge from-layer="4970" from-port="3" to-layer="5018" to-port="0" />
+		<edge from-layer="4970" from-port="3" to-layer="5000" to-port="0" />
 		<edge from-layer="4971" from-port="0" to-layer="4974" to-port="1" />
 		<edge from-layer="4972" from-port="0" to-layer="4974" to-port="2" />
 		<edge from-layer="4973" from-port="0" to-layer="4974" to-port="3" />
-		<edge from-layer="4974" from-port="4" to-layer="4976" to-port="0" />
+		<edge from-layer="4974" from-port="5" to-layer="4976" to-port="0" />
 		<edge from-layer="4975" from-port="0" to-layer="4976" to-port="1" />
-		<edge from-layer="4976" from-port="2" to-layer="4977" to-port="0" />
-		<edge from-layer="4976" from-port="2" to-layer="5069" to-port="0" />
-		<edge from-layer="4976" from-port="2" to-layer="5041" to-port="0" />
-		<edge from-layer="4977" from-port="1" to-layer="4979" to-port="0" />
-		<edge from-layer="4978" from-port="0" to-layer="4979" to-port="2" />
-		<edge from-layer="4979" from-port="3" to-layer="4993" to-port="0" />
-		<edge from-layer="4980" from-port="0" to-layer="4981" to-port="0" />
-		<edge from-layer="4981" from-port="1" to-layer="4986" to-port="0" />
+		<edge from-layer="4976" from-port="2" to-layer="4998" to-port="0" />
+		<edge from-layer="4976" from-port="2" to-layer="4980" to-port="0" />
+		<edge from-layer="4977" from-port="0" to-layer="4980" to-port="1" />
+		<edge from-layer="4978" from-port="0" to-layer="4980" to-port="2" />
+		<edge from-layer="4979" from-port="0" to-layer="4980" to-port="3" />
+		<edge from-layer="4980" from-port="5" to-layer="4985" to-port="0" />
+		<edge from-layer="4980" from-port="5" to-layer="4991" to-port="0" />
+		<edge from-layer="4980" from-port="5" to-layer="4981" to-port="0" />
+		<edge from-layer="4981" from-port="2" to-layer="4994" to-port="0" />
 		<edge from-layer="4982" from-port="0" to-layer="4985" to-port="1" />
 		<edge from-layer="4983" from-port="0" to-layer="4985" to-port="2" />
 		<edge from-layer="4984" from-port="0" to-layer="4985" to-port="3" />
-		<edge from-layer="4985" from-port="4" to-layer="4986" to-port="1" />
-		<edge from-layer="4986" from-port="2" to-layer="4987" to-port="0" />
-		<edge from-layer="4987" from-port="1" to-layer="4989" to-port="0" />
-		<edge from-layer="4988" from-port="0" to-layer="4989" to-port="2" />
-		<edge from-layer="4989" from-port="3" to-layer="4990" to-port="0" />
-		<edge from-layer="4989" from-port="3" to-layer="5071" to-port="0" />
-		<edge from-layer="4990" from-port="1" to-layer="4992" to-port="0" />
-		<edge from-layer="4991" from-port="0" to-layer="4992" to-port="2" />
-		<edge from-layer="4992" from-port="3" to-layer="4993" to-port="1" />
-		<edge from-layer="4993" from-port="2" to-layer="4995" to-port="0" />
-		<edge from-layer="4994" from-port="0" to-layer="4995" to-port="1" />
-		<edge from-layer="4995" from-port="2" to-layer="4997" to-port="2" />
-		<edge from-layer="4995" from-port="2" to-layer="5027" to-port="2" />
-		<edge from-layer="4996" from-port="0" to-layer="4997" to-port="3" />
-		<edge from-layer="4997" from-port="4" to-layer="4999" to-port="0" />
-		<edge from-layer="4998" from-port="0" to-layer="4999" to-port="2" />
-		<edge from-layer="4999" from-port="3" to-layer="5000" to-port="0" />
-		<edge from-layer="5000" from-port="2" to-layer="5042" to-port="1" />
-		<edge from-layer="5000" from-port="2" to-layer="5001" to-port="1" />
-		<edge from-layer="5001" from-port="2" to-layer="5032" to-port="0" />
-		<edge from-layer="5002" from-port="0" to-layer="5012" to-port="0" />
-		<edge from-layer="5003" from-port="0" to-layer="5012" to-port="1" />
-		<edge from-layer="5004" from-port="1" to-layer="5006" to-port="0" />
-		<edge from-layer="5005" from-port="0" to-layer="5006" to-port="2" />
-		<edge from-layer="5006" from-port="3" to-layer="5007" to-port="0" />
-		<edge from-layer="5007" from-port="2" to-layer="5008" to-port="0" />
-		<edge from-layer="5008" from-port="1" to-layer="5010" to-port="0" />
-		<edge from-layer="5009" from-port="0" to-layer="5010" to-port="1" />
-		<edge from-layer="5010" from-port="2" to-layer="5012" to-port="2" />
-		<edge from-layer="5011" from-port="0" to-layer="5012" to-port="3" />
-		<edge from-layer="5012" from-port="4" to-layer="5015" to-port="1" />
-		<edge from-layer="5012" from-port="4" to-layer="5021" to-port="2" />
-		<edge from-layer="5013" from-port="0" to-layer="5015" to-port="2" />
-		<edge from-layer="5014" from-port="0" to-layer="5015" to-port="3" />
-		<edge from-layer="5015" from-port="4" to-layer="5018" to-port="0" />
-		<edge from-layer="5016" from-port="0" to-layer="5017" to-port="0" />
-		<edge from-layer="5017" from-port="1" to-layer="5018" to-port="1" />
-		<edge from-layer="5018" from-port="2" to-layer="5022" to-port="0" />
+		<edge from-layer="4985" from-port="5" to-layer="4987" to-port="0" />
+		<edge from-layer="4986" from-port="0" to-layer="4987" to-port="1" />
+		<edge from-layer="4987" from-port="2" to-layer="4992" to-port="0" />
+		<edge from-layer="4988" from-port="0" to-layer="4991" to-port="1" />
+		<edge from-layer="4989" from-port="0" to-layer="4991" to-port="2" />
+		<edge from-layer="4990" from-port="0" to-layer="4991" to-port="3" />
+		<edge from-layer="4991" from-port="5" to-layer="4992" to-port="1" />
+		<edge from-layer="4992" from-port="2" to-layer="4993" to-port="0" />
+		<edge from-layer="4993" from-port="2" to-layer="4994" to-port="1" />
+		<edge from-layer="4994" from-port="2" to-layer="4999" to-port="0" />
+		<edge from-layer="4995" from-port="0" to-layer="4998" to-port="1" />
+		<edge from-layer="4996" from-port="0" to-layer="4998" to-port="2" />
+		<edge from-layer="4997" from-port="0" to-layer="4998" to-port="3" />
+		<edge from-layer="4998" from-port="5" to-layer="4999" to-port="1" />
+		<edge from-layer="4999" from-port="2" to-layer="5000" to-port="1" />
+		<edge from-layer="5000" from-port="2" to-layer="5111" to-port="0" />
+		<edge from-layer="5000" from-port="2" to-layer="5025" to-port="1" />
+		<edge from-layer="5001" from-port="0" to-layer="5006" to-port="0" />
+		<edge from-layer="5002" from-port="0" to-layer="5005" to-port="1" />
+		<edge from-layer="5003" from-port="0" to-layer="5005" to-port="2" />
+		<edge from-layer="5004" from-port="0" to-layer="5005" to-port="3" />
+		<edge from-layer="5005" from-port="4" to-layer="5006" to-port="1" />
+		<edge from-layer="5006" from-port="2" to-layer="5007" to-port="0" />
+		<edge from-layer="5007" from-port="1" to-layer="5009" to-port="0" />
+		<edge from-layer="5008" from-port="0" to-layer="5009" to-port="2" />
+		<edge from-layer="5009" from-port="3" to-layer="5016" to-port="0" />
+		<edge from-layer="5010" from-port="0" to-layer="5013" to-port="1" />
+		<edge from-layer="5011" from-port="0" to-layer="5013" to-port="2" />
+		<edge from-layer="5012" from-port="0" to-layer="5013" to-port="3" />
+		<edge from-layer="5013" from-port="5" to-layer="5015" to-port="0" />
+		<edge from-layer="5014" from-port="0" to-layer="5015" to-port="1" />
+		<edge from-layer="5015" from-port="2" to-layer="5016" to-port="1" />
+		<edge from-layer="5016" from-port="2" to-layer="5025" to-port="2" />
+		<edge from-layer="5016" from-port="2" to-layer="5110" to-port="0" />
+		<edge from-layer="5017" from-port="0" to-layer="5024" to-port="1" />
+		<edge from-layer="5018" from-port="1" to-layer="5021" to-port="0" />
 		<edge from-layer="5019" from-port="0" to-layer="5021" to-port="1" />
-		<edge from-layer="5020" from-port="0" to-layer="5021" to-port="3" />
-		<edge from-layer="5021" from-port="4" to-layer="5022" to-port="1" />
-		<edge from-layer="5022" from-port="2" to-layer="5031" to-port="0" />
-		<edge from-layer="5023" from-port="0" to-layer="5024" to-port="0" />
-		<edge from-layer="5024" from-port="1" to-layer="5027" to-port="0" />
-		<edge from-layer="5025" from-port="0" to-layer="5027" to-port="1" />
-		<edge from-layer="5026" from-port="0" to-layer="5027" to-port="3" />
-		<edge from-layer="5027" from-port="4" to-layer="5029" to-port="0" />
-		<edge from-layer="5028" from-port="0" to-layer="5029" to-port="2" />
-		<edge from-layer="5029" from-port="3" to-layer="5030" to-port="0" />
-		<edge from-layer="5030" from-port="2" to-layer="5064" to-port="1" />
-		<edge from-layer="5030" from-port="2" to-layer="5031" to-port="1" />
-		<edge from-layer="5031" from-port="2" to-layer="5032" to-port="1" />
-		<edge from-layer="5032" from-port="2" to-layer="5037" to-port="0" />
-		<edge from-layer="5033" from-port="0" to-layer="5036" to-port="1" />
-		<edge from-layer="5034" from-port="0" to-layer="5036" to-port="2" />
-		<edge from-layer="5035" from-port="0" to-layer="5036" to-port="3" />
-		<edge from-layer="5036" from-port="4" to-layer="5037" to-port="1" />
-		<edge from-layer="5037" from-port="2" to-layer="5095" to-port="0" />
-		<edge from-layer="5037" from-port="2" to-layer="5128" to-port="0" />
-		<edge from-layer="5038" from-port="0" to-layer="5041" to-port="1" />
-		<edge from-layer="5039" from-port="0" to-layer="5041" to-port="2" />
-		<edge from-layer="5040" from-port="0" to-layer="5041" to-port="3" />
-		<edge from-layer="5041" from-port="4" to-layer="5045" to-port="0" />
-		<edge from-layer="5041" from-port="4" to-layer="5042" to-port="0" />
-		<edge from-layer="5041" from-port="4" to-layer="5056" to-port="0" />
-		<edge from-layer="5041" from-port="4" to-layer="5062" to-port="0" />
-		<edge from-layer="5042" from-port="2" to-layer="5065" to-port="0" />
-		<edge from-layer="5043" from-port="0" to-layer="5053" to-port="0" />
-		<edge from-layer="5044" from-port="0" to-layer="5053" to-port="1" />
-		<edge from-layer="5045" from-port="1" to-layer="5047" to-port="0" />
-		<edge from-layer="5046" from-port="0" to-layer="5047" to-port="2" />
-		<edge from-layer="5047" from-port="3" to-layer="5048" to-port="0" />
-		<edge from-layer="5048" from-port="2" to-layer="5049" to-port="0" />
-		<edge from-layer="5049" from-port="1" to-layer="5051" to-port="0" />
-		<edge from-layer="5050" from-port="0" to-layer="5051" to-port="1" />
-		<edge from-layer="5051" from-port="2" to-layer="5053" to-port="2" />
-		<edge from-layer="5052" from-port="0" to-layer="5053" to-port="3" />
-		<edge from-layer="5053" from-port="4" to-layer="5056" to-port="1" />
-		<edge from-layer="5053" from-port="4" to-layer="5062" to-port="2" />
-		<edge from-layer="5054" from-port="0" to-layer="5056" to-port="2" />
-		<edge from-layer="5055" from-port="0" to-layer="5056" to-port="3" />
-		<edge from-layer="5056" from-port="4" to-layer="5059" to-port="0" />
-		<edge from-layer="5057" from-port="0" to-layer="5058" to-port="0" />
-		<edge from-layer="5058" from-port="1" to-layer="5059" to-port="1" />
-		<edge from-layer="5059" from-port="2" to-layer="5063" to-port="0" />
-		<edge from-layer="5060" from-port="0" to-layer="5062" to-port="1" />
-		<edge from-layer="5061" from-port="0" to-layer="5062" to-port="3" />
-		<edge from-layer="5062" from-port="4" to-layer="5063" to-port="1" />
-		<edge from-layer="5063" from-port="2" to-layer="5064" to-port="0" />
-		<edge from-layer="5064" from-port="2" to-layer="5065" to-port="1" />
-		<edge from-layer="5065" from-port="2" to-layer="5070" to-port="0" />
-		<edge from-layer="5066" from-port="0" to-layer="5069" to-port="1" />
-		<edge from-layer="5067" from-port="0" to-layer="5069" to-port="2" />
-		<edge from-layer="5068" from-port="0" to-layer="5069" to-port="3" />
-		<edge from-layer="5069" from-port="4" to-layer="5070" to-port="1" />
-		<edge from-layer="5070" from-port="2" to-layer="5071" to-port="1" />
-		<edge from-layer="5071" from-port="2" to-layer="5092" to-port="0" />
-		<edge from-layer="5071" from-port="2" to-layer="7696" to-port="0" />
-		<edge from-layer="5071" from-port="2" to-layer="5128" to-port="1" />
-		<edge from-layer="5072" from-port="0" to-layer="5073" to-port="0" />
-		<edge from-layer="5073" from-port="1" to-layer="5078" to-port="0" />
-		<edge from-layer="5074" from-port="0" to-layer="5077" to-port="1" />
-		<edge from-layer="5075" from-port="0" to-layer="5077" to-port="2" />
-		<edge from-layer="5076" from-port="0" to-layer="5077" to-port="3" />
-		<edge from-layer="5077" from-port="4" to-layer="5078" to-port="1" />
-		<edge from-layer="5078" from-port="2" to-layer="5079" to-port="0" />
-		<edge from-layer="5079" from-port="1" to-layer="5081" to-port="0" />
-		<edge from-layer="5080" from-port="0" to-layer="5081" to-port="2" />
-		<edge from-layer="5081" from-port="3" to-layer="5088" to-port="0" />
-		<edge from-layer="5082" from-port="0" to-layer="5085" to-port="1" />
-		<edge from-layer="5083" from-port="0" to-layer="5085" to-port="2" />
-		<edge from-layer="5084" from-port="0" to-layer="5085" to-port="3" />
-		<edge from-layer="5085" from-port="4" to-layer="5087" to-port="0" />
-		<edge from-layer="5086" from-port="0" to-layer="5087" to-port="1" />
-		<edge from-layer="5087" from-port="2" to-layer="5088" to-port="1" />
-		<edge from-layer="5088" from-port="2" to-layer="7695" to-port="0" />
-		<edge from-layer="5088" from-port="2" to-layer="5128" to-port="2" />
-		<edge from-layer="5089" from-port="0" to-layer="5107" to-port="0" />
-		<edge from-layer="5090" from-port="0" to-layer="5102" to-port="0" />
-		<edge from-layer="5091" from-port="0" to-layer="5102" to-port="1" />
-		<edge from-layer="5091" from-port="0" to-layer="5105" to-port="1" />
-		<edge from-layer="5092" from-port="1" to-layer="5094" to-port="0" />
-		<edge from-layer="5093" from-port="0" to-layer="5094" to-port="2" />
-		<edge from-layer="5094" from-port="3" to-layer="5098" to-port="0" />
-		<edge from-layer="5094" from-port="3" to-layer="5104" to-port="0" />
-		<edge from-layer="5095" from-port="1" to-layer="5123" to-port="0" />
-		<edge from-layer="5095" from-port="1" to-layer="5097" to-port="0" />
-		<edge from-layer="5096" from-port="0" to-layer="5097" to-port="2" />
-		<edge from-layer="5097" from-port="3" to-layer="5098" to-port="1" />
-		<edge from-layer="5098" from-port="2" to-layer="5100" to-port="0" />
-		<edge from-layer="5099" from-port="0" to-layer="5104" to-port="1" />
-		<edge from-layer="5099" from-port="0" to-layer="5100" to-port="1" />
-		<edge from-layer="5100" from-port="2" to-layer="5102" to-port="2" />
-		<edge from-layer="5101" from-port="0" to-layer="5102" to-port="3" />
-		<edge from-layer="5101" from-port="0" to-layer="5105" to-port="3" />
-		<edge from-layer="5102" from-port="4" to-layer="5107" to-port="1" />
-		<edge from-layer="5103" from-port="0" to-layer="5105" to-port="0" />
-		<edge from-layer="5104" from-port="2" to-layer="5105" to-port="2" />
-		<edge from-layer="5104" from-port="2" to-layer="5112" to-port="2" />
-		<edge from-layer="5105" from-port="4" to-layer="5107" to-port="2" />
-		<edge from-layer="5106" from-port="0" to-layer="5107" to-port="3" />
-		<edge from-layer="5107" from-port="4" to-layer="5114" to-port="0" />
-		<edge from-layer="5108" from-port="0" to-layer="5114" to-port="1" />
-		<edge from-layer="5109" from-port="0" to-layer="5112" to-port="0" />
-		<edge from-layer="5110" from-port="0" to-layer="5112" to-port="1" />
-		<edge from-layer="5111" from-port="0" to-layer="5112" to-port="3" />
-		<edge from-layer="5112" from-port="4" to-layer="5114" to-port="2" />
-		<edge from-layer="5113" from-port="0" to-layer="5114" to-port="3" />
-		<edge from-layer="5114" from-port="4" to-layer="5118" to-port="0" />
-		<edge from-layer="5115" from-port="0" to-layer="5116" to-port="0" />
-		<edge from-layer="5116" from-port="1" to-layer="5118" to-port="1" />
-		<edge from-layer="5117" from-port="0" to-layer="5118" to-port="2" />
-		<edge from-layer="5118" from-port="3" to-layer="5119" to-port="0" />
-		<edge from-layer="5119" from-port="2" to-layer="5127" to-port="0" />
-		<edge from-layer="5119" from-port="2" to-layer="5120" to-port="0" />
-		<edge from-layer="5120" from-port="1" to-layer="5126" to-port="0" />
-		<edge from-layer="5121" from-port="0" to-layer="5123" to-port="1" />
-		<edge from-layer="5122" from-port="0" to-layer="5123" to-port="2" />
-		<edge from-layer="5123" from-port="3" to-layer="5124" to-port="0" />
-		<edge from-layer="5124" from-port="4" to-layer="5125" to-port="0" />
-		<edge from-layer="5125" from-port="1" to-layer="5126" to-port="1" />
-		<edge from-layer="5126" from-port="2" to-layer="5127" to-port="1" />
-		<edge from-layer="5127" from-port="2" to-layer="5128" to-port="3" />
-		<edge from-layer="5128" from-port="4" to-layer="5130" to-port="0" />
-		<edge from-layer="5129" from-port="0" to-layer="5130" to-port="1" />
-		<edge from-layer="5130" from-port="2" to-layer="5132" to-port="0" />
-		<edge from-layer="5131" from-port="0" to-layer="5132" to-port="1" />
-		<edge from-layer="5132" from-port="2" to-layer="5141" to-port="0" />
-		<edge from-layer="5133" from-port="0" to-layer="5134" to-port="0" />
-		<edge from-layer="5134" from-port="1" to-layer="5137" to-port="0" />
-		<edge from-layer="5135" from-port="0" to-layer="5136" to-port="0" />
-		<edge from-layer="5136" from-port="1" to-layer="5137" to-port="1" />
-		<edge from-layer="5137" from-port="2" to-layer="5139" to-port="0" />
-		<edge from-layer="5138" from-port="0" to-layer="5139" to-port="1" />
-		<edge from-layer="5139" from-port="2" to-layer="5140" to-port="0" />
-		<edge from-layer="5140" from-port="1" to-layer="5141" to-port="1" />
-		<edge from-layer="5141" from-port="2" to-layer="5144" to-port="0" />
-		<edge from-layer="5142" from-port="0" to-layer="5143" to-port="0" />
-		<edge from-layer="5143" from-port="1" to-layer="5144" to-port="1" />
-		<edge from-layer="5144" from-port="2" to-layer="5145" to-port="1" />
-		<edge from-layer="5145" from-port="2" to-layer="5146" to-port="0" />
-		<edge from-layer="5146" from-port="2" to-layer="5148" to-port="0" />
-		<edge from-layer="5146" from-port="2" to-layer="5388" to-port="1" />
-		<edge from-layer="5147" from-port="0" to-layer="5148" to-port="1" />
-		<edge from-layer="5148" from-port="2" to-layer="5182" to-port="0" />
-		<edge from-layer="5148" from-port="2" to-layer="5151" to-port="0" />
-		<edge from-layer="5149" from-port="0" to-layer="5150" to-port="0" />
-		<edge from-layer="5150" from-port="1" to-layer="5151" to-port="1" />
-		<edge from-layer="5151" from-port="2" to-layer="5154" to-port="0" />
-		<edge from-layer="5152" from-port="0" to-layer="5153" to-port="0" />
-		<edge from-layer="5153" from-port="1" to-layer="5154" to-port="1" />
-		<edge from-layer="5154" from-port="2" to-layer="5163" to-port="0" />
-		<edge from-layer="5155" from-port="0" to-layer="5156" to-port="0" />
-		<edge from-layer="5156" from-port="1" to-layer="5159" to-port="0" />
-		<edge from-layer="5157" from-port="0" to-layer="5158" to-port="0" />
-		<edge from-layer="5158" from-port="1" to-layer="5159" to-port="1" />
-		<edge from-layer="5159" from-port="2" to-layer="5161" to-port="0" />
+		<edge from-layer="5020" from-port="0" to-layer="5021" to-port="2" />
+		<edge from-layer="5021" from-port="3" to-layer="5022" to-port="0" />
+		<edge from-layer="5022" from-port="2" to-layer="5024" to-port="2" />
+		<edge from-layer="5023" from-port="0" to-layer="5024" to-port="3" />
+		<edge from-layer="5024" from-port="5" to-layer="5025" to-port="3" />
+		<edge from-layer="5025" from-port="4" to-layer="5027" to-port="0" />
+		<edge from-layer="5026" from-port="0" to-layer="5027" to-port="1" />
+		<edge from-layer="5027" from-port="2" to-layer="5029" to-port="0" />
+		<edge from-layer="5028" from-port="0" to-layer="5029" to-port="1" />
+		<edge from-layer="5029" from-port="2" to-layer="5038" to-port="0" />
+		<edge from-layer="5030" from-port="0" to-layer="5031" to-port="0" />
+		<edge from-layer="5031" from-port="1" to-layer="5034" to-port="0" />
+		<edge from-layer="5032" from-port="0" to-layer="5033" to-port="0" />
+		<edge from-layer="5033" from-port="1" to-layer="5034" to-port="1" />
+		<edge from-layer="5034" from-port="2" to-layer="5036" to-port="0" />
+		<edge from-layer="5035" from-port="0" to-layer="5036" to-port="1" />
+		<edge from-layer="5036" from-port="2" to-layer="5037" to-port="0" />
+		<edge from-layer="5037" from-port="1" to-layer="5038" to-port="1" />
+		<edge from-layer="5038" from-port="2" to-layer="5040" to-port="0" />
+		<edge from-layer="5039" from-port="0" to-layer="5040" to-port="1" />
+		<edge from-layer="5040" from-port="2" to-layer="5041" to-port="1" />
+		<edge from-layer="5041" from-port="2" to-layer="5042" to-port="0" />
+		<edge from-layer="5042" from-port="2" to-layer="5044" to-port="0" />
+		<edge from-layer="5042" from-port="2" to-layer="5243" to-port="1" />
+		<edge from-layer="5043" from-port="0" to-layer="5044" to-port="1" />
+		<edge from-layer="5044" from-port="2" to-layer="5046" to-port="0" />
+		<edge from-layer="5044" from-port="2" to-layer="5161" to-port="0" />
+		<edge from-layer="5045" from-port="0" to-layer="5046" to-port="1" />
+		<edge from-layer="5046" from-port="2" to-layer="5048" to-port="0" />
+		<edge from-layer="5047" from-port="0" to-layer="5048" to-port="1" />
+		<edge from-layer="5048" from-port="2" to-layer="5057" to-port="0" />
+		<edge from-layer="5049" from-port="0" to-layer="5050" to-port="0" />
+		<edge from-layer="5050" from-port="1" to-layer="5053" to-port="0" />
+		<edge from-layer="5051" from-port="0" to-layer="5052" to-port="0" />
+		<edge from-layer="5052" from-port="1" to-layer="5053" to-port="1" />
+		<edge from-layer="5053" from-port="2" to-layer="5055" to-port="0" />
+		<edge from-layer="5054" from-port="0" to-layer="5055" to-port="1" />
+		<edge from-layer="5055" from-port="2" to-layer="5056" to-port="0" />
+		<edge from-layer="5056" from-port="1" to-layer="5057" to-port="1" />
+		<edge from-layer="5057" from-port="2" to-layer="5059" to-port="0" />
+		<edge from-layer="5058" from-port="0" to-layer="5059" to-port="1" />
+		<edge from-layer="5059" from-port="2" to-layer="5061" to-port="0" />
+		<edge from-layer="5060" from-port="0" to-layer="5061" to-port="1" />
+		<edge from-layer="5061" from-port="2" to-layer="5192" to-port="0" />
+		<edge from-layer="5061" from-port="2" to-layer="5065" to-port="0" />
+		<edge from-layer="5061" from-port="2" to-layer="5082" to-port="0" />
+		<edge from-layer="5062" from-port="0" to-layer="5065" to-port="1" />
+		<edge from-layer="5063" from-port="0" to-layer="5065" to-port="2" />
+		<edge from-layer="5064" from-port="0" to-layer="5065" to-port="3" />
+		<edge from-layer="5065" from-port="5" to-layer="5067" to-port="0" />
+		<edge from-layer="5066" from-port="0" to-layer="5067" to-port="1" />
+		<edge from-layer="5067" from-port="2" to-layer="5068" to-port="1" />
+		<edge from-layer="5068" from-port="2" to-layer="5226" to-port="2" />
+		<edge from-layer="5068" from-port="2" to-layer="5069" to-port="0" />
+		<edge from-layer="5070" from-port="0" to-layer="5075" to-port="0" />
+		<edge from-layer="5071" from-port="0" to-layer="5074" to-port="1" />
+		<edge from-layer="5072" from-port="0" to-layer="5074" to-port="2" />
+		<edge from-layer="5073" from-port="0" to-layer="5074" to-port="3" />
+		<edge from-layer="5074" from-port="4" to-layer="5075" to-port="1" />
+		<edge from-layer="5075" from-port="2" to-layer="5076" to-port="0" />
+		<edge from-layer="5076" from-port="1" to-layer="5078" to-port="0" />
+		<edge from-layer="5077" from-port="0" to-layer="5078" to-port="2" />
+		<edge from-layer="5078" from-port="3" to-layer="5219" to-port="0" />
+		<edge from-layer="5078" from-port="3" to-layer="5108" to-port="0" />
+		<edge from-layer="5079" from-port="0" to-layer="5082" to-port="1" />
+		<edge from-layer="5080" from-port="0" to-layer="5082" to-port="2" />
+		<edge from-layer="5081" from-port="0" to-layer="5082" to-port="3" />
+		<edge from-layer="5082" from-port="5" to-layer="5084" to-port="0" />
+		<edge from-layer="5083" from-port="0" to-layer="5084" to-port="1" />
+		<edge from-layer="5084" from-port="2" to-layer="5088" to-port="0" />
+		<edge from-layer="5084" from-port="2" to-layer="5106" to-port="0" />
+		<edge from-layer="5085" from-port="0" to-layer="5088" to-port="1" />
+		<edge from-layer="5086" from-port="0" to-layer="5088" to-port="2" />
+		<edge from-layer="5087" from-port="0" to-layer="5088" to-port="3" />
+		<edge from-layer="5088" from-port="5" to-layer="5089" to-port="0" />
+		<edge from-layer="5088" from-port="5" to-layer="5093" to-port="0" />
+		<edge from-layer="5088" from-port="5" to-layer="5099" to-port="0" />
+		<edge from-layer="5089" from-port="2" to-layer="5102" to-port="0" />
+		<edge from-layer="5090" from-port="0" to-layer="5093" to-port="1" />
+		<edge from-layer="5091" from-port="0" to-layer="5093" to-port="2" />
+		<edge from-layer="5092" from-port="0" to-layer="5093" to-port="3" />
+		<edge from-layer="5093" from-port="5" to-layer="5095" to-port="0" />
+		<edge from-layer="5094" from-port="0" to-layer="5095" to-port="1" />
+		<edge from-layer="5095" from-port="2" to-layer="5100" to-port="0" />
+		<edge from-layer="5096" from-port="0" to-layer="5099" to-port="1" />
+		<edge from-layer="5097" from-port="0" to-layer="5099" to-port="2" />
+		<edge from-layer="5098" from-port="0" to-layer="5099" to-port="3" />
+		<edge from-layer="5099" from-port="5" to-layer="5100" to-port="1" />
+		<edge from-layer="5100" from-port="2" to-layer="5101" to-port="0" />
+		<edge from-layer="5101" from-port="2" to-layer="5102" to-port="1" />
+		<edge from-layer="5102" from-port="2" to-layer="5107" to-port="0" />
+		<edge from-layer="5103" from-port="0" to-layer="5106" to-port="1" />
+		<edge from-layer="5104" from-port="0" to-layer="5106" to-port="2" />
+		<edge from-layer="5105" from-port="0" to-layer="5106" to-port="3" />
+		<edge from-layer="5106" from-port="5" to-layer="5107" to-port="1" />
+		<edge from-layer="5107" from-port="2" to-layer="5108" to-port="1" />
+		<edge from-layer="5108" from-port="2" to-layer="5109" to-port="0" />
+		<edge from-layer="5108" from-port="2" to-layer="5226" to-port="1" />
 		<edge from-layer="5160" from-port="0" to-layer="5161" to-port="1" />
-		<edge from-layer="5161" from-port="2" to-layer="5162" to-port="0" />
-		<edge from-layer="5162" from-port="1" to-layer="5163" to-port="1" />
-		<edge from-layer="5163" from-port="2" to-layer="5166" to-port="0" />
+		<edge from-layer="5161" from-port="2" to-layer="5163" to-port="0" />
+		<edge from-layer="5162" from-port="0" to-layer="5163" to-port="1" />
+		<edge from-layer="5163" from-port="2" to-layer="5172" to-port="0" />
 		<edge from-layer="5164" from-port="0" to-layer="5165" to-port="0" />
-		<edge from-layer="5165" from-port="1" to-layer="5166" to-port="1" />
-		<edge from-layer="5166" from-port="2" to-layer="5167" to-port="0" />
-		<edge from-layer="5167" from-port="1" to-layer="5176" to-port="0" />
-		<edge from-layer="5168" from-port="0" to-layer="5169" to-port="0" />
-		<edge from-layer="5169" from-port="1" to-layer="5172" to-port="0" />
-		<edge from-layer="5170" from-port="0" to-layer="5171" to-port="0" />
+		<edge from-layer="5165" from-port="1" to-layer="5168" to-port="0" />
+		<edge from-layer="5166" from-port="0" to-layer="5167" to-port="0" />
+		<edge from-layer="5167" from-port="1" to-layer="5168" to-port="1" />
+		<edge from-layer="5168" from-port="2" to-layer="5170" to-port="0" />
+		<edge from-layer="5169" from-port="0" to-layer="5170" to-port="1" />
+		<edge from-layer="5170" from-port="2" to-layer="5171" to-port="0" />
 		<edge from-layer="5171" from-port="1" to-layer="5172" to-port="1" />
-		<edge from-layer="5172" from-port="2" to-layer="5174" to-port="0" />
-		<edge from-layer="5173" from-port="0" to-layer="5174" to-port="1" />
-		<edge from-layer="5174" from-port="2" to-layer="5175" to-port="0" />
-		<edge from-layer="5175" from-port="1" to-layer="5176" to-port="1" />
-		<edge from-layer="5176" from-port="2" to-layer="5179" to-port="0" />
+		<edge from-layer="5172" from-port="2" to-layer="5175" to-port="0" />
+		<edge from-layer="5173" from-port="0" to-layer="5174" to-port="0" />
+		<edge from-layer="5174" from-port="1" to-layer="5175" to-port="1" />
+		<edge from-layer="5175" from-port="2" to-layer="5176" to-port="0" />
+		<edge from-layer="5176" from-port="1" to-layer="5185" to-port="0" />
 		<edge from-layer="5177" from-port="0" to-layer="5178" to-port="0" />
-		<edge from-layer="5178" from-port="1" to-layer="5179" to-port="1" />
-		<edge from-layer="5179" from-port="2" to-layer="5387" to-port="0" />
-		<edge from-layer="5180" from-port="0" to-layer="5181" to-port="0" />
-		<edge from-layer="5181" from-port="1" to-layer="5182" to-port="1" />
-		<edge from-layer="5182" from-port="2" to-layer="5185" to-port="0" />
-		<edge from-layer="5183" from-port="0" to-layer="5184" to-port="0" />
+		<edge from-layer="5178" from-port="1" to-layer="5181" to-port="0" />
+		<edge from-layer="5179" from-port="0" to-layer="5180" to-port="0" />
+		<edge from-layer="5180" from-port="1" to-layer="5181" to-port="1" />
+		<edge from-layer="5181" from-port="2" to-layer="5183" to-port="0" />
+		<edge from-layer="5182" from-port="0" to-layer="5183" to-port="1" />
+		<edge from-layer="5183" from-port="2" to-layer="5184" to-port="0" />
 		<edge from-layer="5184" from-port="1" to-layer="5185" to-port="1" />
-		<edge from-layer="5185" from-port="2" to-layer="5194" to-port="0" />
+		<edge from-layer="5185" from-port="2" to-layer="5188" to-port="0" />
 		<edge from-layer="5186" from-port="0" to-layer="5187" to-port="0" />
-		<edge from-layer="5187" from-port="1" to-layer="5190" to-port="0" />
-		<edge from-layer="5188" from-port="0" to-layer="5189" to-port="0" />
-		<edge from-layer="5189" from-port="1" to-layer="5190" to-port="1" />
-		<edge from-layer="5190" from-port="2" to-layer="5192" to-port="0" />
-		<edge from-layer="5191" from-port="0" to-layer="5192" to-port="1" />
-		<edge from-layer="5192" from-port="2" to-layer="5193" to-port="0" />
-		<edge from-layer="5193" from-port="1" to-layer="5194" to-port="1" />
-		<edge from-layer="5194" from-port="2" to-layer="5197" to-port="0" />
-		<edge from-layer="5195" from-port="0" to-layer="5196" to-port="0" />
-		<edge from-layer="5196" from-port="1" to-layer="5197" to-port="1" />
-		<edge from-layer="5197" from-port="2" to-layer="5199" to-port="0" />
-		<edge from-layer="5198" from-port="0" to-layer="5199" to-port="1" />
-		<edge from-layer="5199" from-port="2" to-layer="5203" to-port="0" />
-		<edge from-layer="5199" from-port="2" to-layer="5327" to-port="0" />
-		<edge from-layer="5199" from-port="2" to-layer="5216" to-port="0" />
+		<edge from-layer="5187" from-port="1" to-layer="5188" to-port="1" />
+		<edge from-layer="5188" from-port="2" to-layer="5242" to-port="0" />
+		<edge from-layer="5189" from-port="0" to-layer="5192" to-port="1" />
+		<edge from-layer="5190" from-port="0" to-layer="5192" to-port="2" />
+		<edge from-layer="5191" from-port="0" to-layer="5192" to-port="3" />
+		<edge from-layer="5192" from-port="5" to-layer="5194" to-port="0" />
+		<edge from-layer="5193" from-port="0" to-layer="5194" to-port="1" />
+		<edge from-layer="5194" from-port="2" to-layer="5216" to-port="0" />
+		<edge from-layer="5194" from-port="2" to-layer="5198" to-port="0" />
+		<edge from-layer="5195" from-port="0" to-layer="5198" to-port="1" />
+		<edge from-layer="5196" from-port="0" to-layer="5198" to-port="2" />
+		<edge from-layer="5197" from-port="0" to-layer="5198" to-port="3" />
+		<edge from-layer="5198" from-port="5" to-layer="5209" to-port="0" />
+		<edge from-layer="5198" from-port="5" to-layer="5203" to-port="0" />
+		<edge from-layer="5198" from-port="5" to-layer="5199" to-port="0" />
+		<edge from-layer="5199" from-port="2" to-layer="5212" to-port="0" />
 		<edge from-layer="5200" from-port="0" to-layer="5203" to-port="1" />
 		<edge from-layer="5201" from-port="0" to-layer="5203" to-port="2" />
 		<edge from-layer="5202" from-port="0" to-layer="5203" to-port="3" />
-		<edge from-layer="5203" from-port="4" to-layer="5205" to-port="0" />
+		<edge from-layer="5203" from-port="5" to-layer="5205" to-port="0" />
 		<edge from-layer="5204" from-port="0" to-layer="5205" to-port="1" />
-		<edge from-layer="5205" from-port="2" to-layer="5209" to-port="0" />
-		<edge from-layer="5205" from-port="2" to-layer="5278" to-port="0" />
+		<edge from-layer="5205" from-port="2" to-layer="5210" to-port="0" />
 		<edge from-layer="5206" from-port="0" to-layer="5209" to-port="1" />
 		<edge from-layer="5207" from-port="0" to-layer="5209" to-port="2" />
 		<edge from-layer="5208" from-port="0" to-layer="5209" to-port="3" />
-		<edge from-layer="5209" from-port="4" to-layer="5257" to-port="0" />
-		<edge from-layer="5209" from-port="4" to-layer="5246" to-port="0" />
-		<edge from-layer="5209" from-port="4" to-layer="5243" to-port="0" />
-		<edge from-layer="5209" from-port="4" to-layer="5263" to-port="0" />
-		<edge from-layer="5210" from-port="0" to-layer="5211" to-port="0" />
-		<edge from-layer="5211" from-port="1" to-layer="5239" to-port="0" />
-		<edge from-layer="5212" from-port="0" to-layer="5239" to-port="1" />
+		<edge from-layer="5209" from-port="5" to-layer="5210" to-port="1" />
+		<edge from-layer="5210" from-port="2" to-layer="5211" to-port="0" />
+		<edge from-layer="5211" from-port="2" to-layer="5212" to-port="1" />
+		<edge from-layer="5212" from-port="2" to-layer="5217" to-port="0" />
 		<edge from-layer="5213" from-port="0" to-layer="5216" to-port="1" />
 		<edge from-layer="5214" from-port="0" to-layer="5216" to-port="2" />
 		<edge from-layer="5215" from-port="0" to-layer="5216" to-port="3" />
-		<edge from-layer="5216" from-port="4" to-layer="5218" to-port="0" />
-		<edge from-layer="5217" from-port="0" to-layer="5218" to-port="1" />
-		<edge from-layer="5218" from-port="2" to-layer="5219" to-port="0" />
-		<edge from-layer="5218" from-port="2" to-layer="5283" to-port="0" />
-		<edge from-layer="5218" from-port="2" to-layer="5311" to-port="0" />
-		<edge from-layer="5219" from-port="1" to-layer="5221" to-port="0" />
-		<edge from-layer="5220" from-port="0" to-layer="5221" to-port="2" />
-		<edge from-layer="5221" from-port="3" to-layer="5235" to-port="0" />
-		<edge from-layer="5222" from-port="0" to-layer="5223" to-port="0" />
-		<edge from-layer="5223" from-port="1" to-layer="5228" to-port="0" />
-		<edge from-layer="5224" from-port="0" to-layer="5227" to-port="1" />
-		<edge from-layer="5225" from-port="0" to-layer="5227" to-port="2" />
-		<edge from-layer="5226" from-port="0" to-layer="5227" to-port="3" />
-		<edge from-layer="5227" from-port="4" to-layer="5228" to-port="1" />
-		<edge from-layer="5228" from-port="2" to-layer="5229" to-port="0" />
-		<edge from-layer="5229" from-port="1" to-layer="5231" to-port="0" />
-		<edge from-layer="5230" from-port="0" to-layer="5231" to-port="2" />
-		<edge from-layer="5231" from-port="3" to-layer="5232" to-port="0" />
-		<edge from-layer="5231" from-port="3" to-layer="5313" to-port="0" />
-		<edge from-layer="5232" from-port="1" to-layer="5234" to-port="0" />
-		<edge from-layer="5233" from-port="0" to-layer="5234" to-port="2" />
-		<edge from-layer="5234" from-port="3" to-layer="5235" to-port="1" />
+		<edge from-layer="5216" from-port="5" to-layer="5217" to-port="1" />
+		<edge from-layer="5217" from-port="2" to-layer="5226" to-port="0" />
+		<edge from-layer="5218" from-port="0" to-layer="5225" to-port="1" />
+		<edge from-layer="5219" from-port="1" to-layer="5222" to-port="0" />
+		<edge from-layer="5220" from-port="0" to-layer="5222" to-port="1" />
+		<edge from-layer="5221" from-port="0" to-layer="5222" to-port="2" />
+		<edge from-layer="5222" from-port="3" to-layer="5223" to-port="0" />
+		<edge from-layer="5223" from-port="2" to-layer="5225" to-port="2" />
+		<edge from-layer="5224" from-port="0" to-layer="5225" to-port="3" />
+		<edge from-layer="5225" from-port="5" to-layer="5226" to-port="3" />
+		<edge from-layer="5226" from-port="4" to-layer="5228" to-port="0" />
+		<edge from-layer="5227" from-port="0" to-layer="5228" to-port="1" />
+		<edge from-layer="5228" from-port="2" to-layer="5230" to-port="0" />
+		<edge from-layer="5229" from-port="0" to-layer="5230" to-port="1" />
+		<edge from-layer="5230" from-port="2" to-layer="5239" to-port="0" />
+		<edge from-layer="5231" from-port="0" to-layer="5232" to-port="0" />
+		<edge from-layer="5232" from-port="1" to-layer="5235" to-port="0" />
+		<edge from-layer="5233" from-port="0" to-layer="5234" to-port="0" />
+		<edge from-layer="5234" from-port="1" to-layer="5235" to-port="1" />
 		<edge from-layer="5235" from-port="2" to-layer="5237" to-port="0" />
 		<edge from-layer="5236" from-port="0" to-layer="5237" to-port="1" />
-		<edge from-layer="5237" from-port="2" to-layer="5269" to-port="2" />
-		<edge from-layer="5237" from-port="2" to-layer="5239" to-port="2" />
-		<edge from-layer="5238" from-port="0" to-layer="5239" to-port="3" />
-		<edge from-layer="5239" from-port="4" to-layer="5241" to-port="0" />
-		<edge from-layer="5240" from-port="0" to-layer="5241" to-port="2" />
-		<edge from-layer="5241" from-port="3" to-layer="5242" to-port="0" />
-		<edge from-layer="5242" from-port="2" to-layer="5284" to-port="1" />
-		<edge from-layer="5242" from-port="2" to-layer="5243" to-port="1" />
-		<edge from-layer="5243" from-port="2" to-layer="5274" to-port="0" />
-		<edge from-layer="5244" from-port="0" to-layer="5254" to-port="0" />
-		<edge from-layer="5245" from-port="0" to-layer="5254" to-port="1" />
-		<edge from-layer="5246" from-port="1" to-layer="5248" to-port="0" />
-		<edge from-layer="5247" from-port="0" to-layer="5248" to-port="2" />
-		<edge from-layer="5248" from-port="3" to-layer="5249" to-port="0" />
-		<edge from-layer="5249" from-port="2" to-layer="5250" to-port="0" />
-		<edge from-layer="5250" from-port="1" to-layer="5252" to-port="0" />
-		<edge from-layer="5251" from-port="0" to-layer="5252" to-port="1" />
-		<edge from-layer="5252" from-port="2" to-layer="5254" to-port="2" />
-		<edge from-layer="5253" from-port="0" to-layer="5254" to-port="3" />
-		<edge from-layer="5254" from-port="4" to-layer="5257" to-port="1" />
-		<edge from-layer="5254" from-port="4" to-layer="5263" to-port="2" />
-		<edge from-layer="5255" from-port="0" to-layer="5257" to-port="2" />
-		<edge from-layer="5256" from-port="0" to-layer="5257" to-port="3" />
-		<edge from-layer="5257" from-port="4" to-layer="5260" to-port="0" />
-		<edge from-layer="5258" from-port="0" to-layer="5259" to-port="0" />
-		<edge from-layer="5259" from-port="1" to-layer="5260" to-port="1" />
-		<edge from-layer="5260" from-port="2" to-layer="5264" to-port="0" />
-		<edge from-layer="5261" from-port="0" to-layer="5263" to-port="1" />
-		<edge from-layer="5262" from-port="0" to-layer="5263" to-port="3" />
-		<edge from-layer="5263" from-port="4" to-layer="5264" to-port="1" />
-		<edge from-layer="5264" from-port="2" to-layer="5273" to-port="0" />
-		<edge from-layer="5265" from-port="0" to-layer="5266" to-port="0" />
-		<edge from-layer="5266" from-port="1" to-layer="5269" to-port="0" />
-		<edge from-layer="5267" from-port="0" to-layer="5269" to-port="1" />
-		<edge from-layer="5268" from-port="0" to-layer="5269" to-port="3" />
-		<edge from-layer="5269" from-port="4" to-layer="5271" to-port="0" />
-		<edge from-layer="5270" from-port="0" to-layer="5271" to-port="2" />
-		<edge from-layer="5271" from-port="3" to-layer="5272" to-port="0" />
-		<edge from-layer="5272" from-port="2" to-layer="5273" to-port="1" />
-		<edge from-layer="5272" from-port="2" to-layer="5306" to-port="1" />
-		<edge from-layer="5273" from-port="2" to-layer="5274" to-port="1" />
-		<edge from-layer="5274" from-port="2" to-layer="5279" to-port="0" />
-		<edge from-layer="5275" from-port="0" to-layer="5278" to-port="1" />
-		<edge from-layer="5276" from-port="0" to-layer="5278" to-port="2" />
-		<edge from-layer="5277" from-port="0" to-layer="5278" to-port="3" />
-		<edge from-layer="5278" from-port="4" to-layer="5279" to-port="1" />
-		<edge from-layer="5279" from-port="2" to-layer="5370" to-port="0" />
-		<edge from-layer="5279" from-port="2" to-layer="5337" to-port="0" />
-		<edge from-layer="5280" from-port="0" to-layer="5283" to-port="1" />
-		<edge from-layer="5281" from-port="0" to-layer="5283" to-port="2" />
-		<edge from-layer="5282" from-port="0" to-layer="5283" to-port="3" />
-		<edge from-layer="5283" from-port="4" to-layer="5284" to-port="0" />
-		<edge from-layer="5283" from-port="4" to-layer="5287" to-port="0" />
-		<edge from-layer="5283" from-port="4" to-layer="5304" to-port="0" />
-		<edge from-layer="5283" from-port="4" to-layer="5298" to-port="0" />
-		<edge from-layer="5284" from-port="2" to-layer="5307" to-port="0" />
-		<edge from-layer="5285" from-port="0" to-layer="5295" to-port="0" />
-		<edge from-layer="5286" from-port="0" to-layer="5295" to-port="1" />
-		<edge from-layer="5287" from-port="1" to-layer="5289" to-port="0" />
-		<edge from-layer="5288" from-port="0" to-layer="5289" to-port="2" />
-		<edge from-layer="5289" from-port="3" to-layer="5290" to-port="0" />
-		<edge from-layer="5290" from-port="2" to-layer="5291" to-port="0" />
-		<edge from-layer="5291" from-port="1" to-layer="5293" to-port="0" />
-		<edge from-layer="5292" from-port="0" to-layer="5293" to-port="1" />
-		<edge from-layer="5293" from-port="2" to-layer="5295" to-port="2" />
-		<edge from-layer="5294" from-port="0" to-layer="5295" to-port="3" />
-		<edge from-layer="5295" from-port="4" to-layer="5298" to-port="1" />
-		<edge from-layer="5295" from-port="4" to-layer="5304" to-port="2" />
-		<edge from-layer="5296" from-port="0" to-layer="5298" to-port="2" />
-		<edge from-layer="5297" from-port="0" to-layer="5298" to-port="3" />
-		<edge from-layer="5298" from-port="4" to-layer="5301" to-port="0" />
-		<edge from-layer="5299" from-port="0" to-layer="5300" to-port="0" />
-		<edge from-layer="5300" from-port="1" to-layer="5301" to-port="1" />
-		<edge from-layer="5301" from-port="2" to-layer="5305" to-port="0" />
-		<edge from-layer="5302" from-port="0" to-layer="5304" to-port="1" />
-		<edge from-layer="5303" from-port="0" to-layer="5304" to-port="3" />
-		<edge from-layer="5304" from-port="4" to-layer="5305" to-port="1" />
-		<edge from-layer="5305" from-port="2" to-layer="5306" to-port="0" />
-		<edge from-layer="5306" from-port="2" to-layer="5307" to-port="1" />
-		<edge from-layer="5307" from-port="2" to-layer="5312" to-port="0" />
-		<edge from-layer="5308" from-port="0" to-layer="5311" to-port="1" />
-		<edge from-layer="5309" from-port="0" to-layer="5311" to-port="2" />
-		<edge from-layer="5310" from-port="0" to-layer="5311" to-port="3" />
-		<edge from-layer="5311" from-port="4" to-layer="5312" to-port="1" />
-		<edge from-layer="5312" from-port="2" to-layer="5313" to-port="1" />
-		<edge from-layer="5313" from-port="2" to-layer="5334" to-port="0" />
-		<edge from-layer="5313" from-port="2" to-layer="7694" to-port="0" />
-		<edge from-layer="5313" from-port="2" to-layer="5370" to-port="1" />
-		<edge from-layer="5314" from-port="0" to-layer="5315" to-port="0" />
-		<edge from-layer="5315" from-port="1" to-layer="5320" to-port="0" />
-		<edge from-layer="5316" from-port="0" to-layer="5319" to-port="1" />
-		<edge from-layer="5317" from-port="0" to-layer="5319" to-port="2" />
-		<edge from-layer="5318" from-port="0" to-layer="5319" to-port="3" />
-		<edge from-layer="5319" from-port="4" to-layer="5320" to-port="1" />
-		<edge from-layer="5320" from-port="2" to-layer="5321" to-port="0" />
-		<edge from-layer="5321" from-port="1" to-layer="5323" to-port="0" />
-		<edge from-layer="5322" from-port="0" to-layer="5323" to-port="2" />
-		<edge from-layer="5323" from-port="3" to-layer="5330" to-port="0" />
-		<edge from-layer="5324" from-port="0" to-layer="5327" to-port="1" />
-		<edge from-layer="5325" from-port="0" to-layer="5327" to-port="2" />
-		<edge from-layer="5326" from-port="0" to-layer="5327" to-port="3" />
-		<edge from-layer="5327" from-port="4" to-layer="5329" to-port="0" />
-		<edge from-layer="5328" from-port="0" to-layer="5329" to-port="1" />
-		<edge from-layer="5329" from-port="2" to-layer="5330" to-port="1" />
-		<edge from-layer="5330" from-port="2" to-layer="5370" to-port="2" />
-		<edge from-layer="5330" from-port="2" to-layer="7693" to-port="0" />
-		<edge from-layer="5331" from-port="0" to-layer="5349" to-port="0" />
-		<edge from-layer="5332" from-port="0" to-layer="5344" to-port="0" />
-		<edge from-layer="5333" from-port="0" to-layer="5344" to-port="1" />
-		<edge from-layer="5333" from-port="0" to-layer="5347" to-port="1" />
-		<edge from-layer="5334" from-port="1" to-layer="5336" to-port="0" />
-		<edge from-layer="5335" from-port="0" to-layer="5336" to-port="2" />
-		<edge from-layer="5336" from-port="3" to-layer="5340" to-port="0" />
-		<edge from-layer="5336" from-port="3" to-layer="5346" to-port="0" />
-		<edge from-layer="5337" from-port="1" to-layer="5365" to-port="0" />
-		<edge from-layer="5337" from-port="1" to-layer="5339" to-port="0" />
-		<edge from-layer="5338" from-port="0" to-layer="5339" to-port="2" />
-		<edge from-layer="5339" from-port="3" to-layer="5340" to-port="1" />
-		<edge from-layer="5340" from-port="2" to-layer="5342" to-port="0" />
-		<edge from-layer="5341" from-port="0" to-layer="5346" to-port="1" />
-		<edge from-layer="5341" from-port="0" to-layer="5342" to-port="1" />
-		<edge from-layer="5342" from-port="2" to-layer="5344" to-port="2" />
-		<edge from-layer="5343" from-port="0" to-layer="5344" to-port="3" />
-		<edge from-layer="5343" from-port="0" to-layer="5347" to-port="3" />
-		<edge from-layer="5344" from-port="4" to-layer="5349" to-port="1" />
-		<edge from-layer="5345" from-port="0" to-layer="5347" to-port="0" />
-		<edge from-layer="5346" from-port="2" to-layer="5347" to-port="2" />
-		<edge from-layer="5346" from-port="2" to-layer="5354" to-port="2" />
-		<edge from-layer="5347" from-port="4" to-layer="5349" to-port="2" />
-		<edge from-layer="5348" from-port="0" to-layer="5349" to-port="3" />
-		<edge from-layer="5349" from-port="4" to-layer="5356" to-port="0" />
-		<edge from-layer="5350" from-port="0" to-layer="5356" to-port="1" />
-		<edge from-layer="5351" from-port="0" to-layer="5354" to-port="0" />
-		<edge from-layer="5352" from-port="0" to-layer="5354" to-port="1" />
-		<edge from-layer="5353" from-port="0" to-layer="5354" to-port="3" />
-		<edge from-layer="5354" from-port="4" to-layer="5356" to-port="2" />
-		<edge from-layer="5355" from-port="0" to-layer="5356" to-port="3" />
-		<edge from-layer="5356" from-port="4" to-layer="5360" to-port="0" />
-		<edge from-layer="5357" from-port="0" to-layer="5358" to-port="0" />
-		<edge from-layer="5358" from-port="1" to-layer="5360" to-port="1" />
-		<edge from-layer="5359" from-port="0" to-layer="5360" to-port="2" />
-		<edge from-layer="5360" from-port="3" to-layer="5361" to-port="0" />
-		<edge from-layer="5361" from-port="2" to-layer="5369" to-port="0" />
-		<edge from-layer="5361" from-port="2" to-layer="5362" to-port="0" />
-		<edge from-layer="5362" from-port="1" to-layer="5368" to-port="0" />
-		<edge from-layer="5363" from-port="0" to-layer="5365" to-port="1" />
-		<edge from-layer="5364" from-port="0" to-layer="5365" to-port="2" />
-		<edge from-layer="5365" from-port="3" to-layer="5366" to-port="0" />
-		<edge from-layer="5366" from-port="4" to-layer="5367" to-port="0" />
-		<edge from-layer="5367" from-port="1" to-layer="5368" to-port="1" />
-		<edge from-layer="5368" from-port="2" to-layer="5369" to-port="1" />
-		<edge from-layer="5369" from-port="2" to-layer="5370" to-port="3" />
-		<edge from-layer="5370" from-port="4" to-layer="5372" to-port="0" />
-		<edge from-layer="5371" from-port="0" to-layer="5372" to-port="1" />
-		<edge from-layer="5372" from-port="2" to-layer="5374" to-port="0" />
-		<edge from-layer="5373" from-port="0" to-layer="5374" to-port="1" />
-		<edge from-layer="5374" from-port="2" to-layer="5383" to-port="0" />
-		<edge from-layer="5375" from-port="0" to-layer="5376" to-port="0" />
-		<edge from-layer="5376" from-port="1" to-layer="5379" to-port="0" />
-		<edge from-layer="5377" from-port="0" to-layer="5378" to-port="0" />
-		<edge from-layer="5378" from-port="1" to-layer="5379" to-port="1" />
-		<edge from-layer="5379" from-port="2" to-layer="5381" to-port="0" />
-		<edge from-layer="5380" from-port="0" to-layer="5381" to-port="1" />
-		<edge from-layer="5381" from-port="2" to-layer="5382" to-port="0" />
-		<edge from-layer="5382" from-port="1" to-layer="5383" to-port="1" />
-		<edge from-layer="5383" from-port="2" to-layer="5386" to-port="0" />
-		<edge from-layer="5384" from-port="0" to-layer="5385" to-port="0" />
-		<edge from-layer="5385" from-port="1" to-layer="5386" to-port="1" />
-		<edge from-layer="5386" from-port="2" to-layer="5387" to-port="1" />
-		<edge from-layer="5387" from-port="2" to-layer="5388" to-port="0" />
-		<edge from-layer="5388" from-port="2" to-layer="5390" to-port="0" />
-		<edge from-layer="5388" from-port="2" to-layer="5630" to-port="1" />
-		<edge from-layer="5389" from-port="0" to-layer="5390" to-port="1" />
-		<edge from-layer="5390" from-port="2" to-layer="5393" to-port="0" />
-		<edge from-layer="5390" from-port="2" to-layer="5424" to-port="0" />
-		<edge from-layer="5391" from-port="0" to-layer="5392" to-port="0" />
-		<edge from-layer="5392" from-port="1" to-layer="5393" to-port="1" />
-		<edge from-layer="5393" from-port="2" to-layer="5396" to-port="0" />
-		<edge from-layer="5394" from-port="0" to-layer="5395" to-port="0" />
-		<edge from-layer="5395" from-port="1" to-layer="5396" to-port="1" />
-		<edge from-layer="5396" from-port="2" to-layer="5405" to-port="0" />
-		<edge from-layer="5397" from-port="0" to-layer="5398" to-port="0" />
-		<edge from-layer="5398" from-port="1" to-layer="5401" to-port="0" />
-		<edge from-layer="5399" from-port="0" to-layer="5400" to-port="0" />
-		<edge from-layer="5400" from-port="1" to-layer="5401" to-port="1" />
-		<edge from-layer="5401" from-port="2" to-layer="5403" to-port="0" />
-		<edge from-layer="5402" from-port="0" to-layer="5403" to-port="1" />
-		<edge from-layer="5403" from-port="2" to-layer="5404" to-port="0" />
-		<edge from-layer="5404" from-port="1" to-layer="5405" to-port="1" />
-		<edge from-layer="5405" from-port="2" to-layer="5408" to-port="0" />
-		<edge from-layer="5406" from-port="0" to-layer="5407" to-port="0" />
-		<edge from-layer="5407" from-port="1" to-layer="5408" to-port="1" />
-		<edge from-layer="5408" from-port="2" to-layer="5409" to-port="0" />
-		<edge from-layer="5409" from-port="1" to-layer="5418" to-port="0" />
-		<edge from-layer="5410" from-port="0" to-layer="5411" to-port="0" />
-		<edge from-layer="5411" from-port="1" to-layer="5414" to-port="0" />
-		<edge from-layer="5412" from-port="0" to-layer="5413" to-port="0" />
-		<edge from-layer="5413" from-port="1" to-layer="5414" to-port="1" />
-		<edge from-layer="5414" from-port="2" to-layer="5416" to-port="0" />
-		<edge from-layer="5415" from-port="0" to-layer="5416" to-port="1" />
-		<edge from-layer="5416" from-port="2" to-layer="5417" to-port="0" />
-		<edge from-layer="5417" from-port="1" to-layer="5418" to-port="1" />
-		<edge from-layer="5418" from-port="2" to-layer="5421" to-port="0" />
-		<edge from-layer="5419" from-port="0" to-layer="5420" to-port="0" />
-		<edge from-layer="5420" from-port="1" to-layer="5421" to-port="1" />
-		<edge from-layer="5421" from-port="2" to-layer="5629" to-port="0" />
-		<edge from-layer="5422" from-port="0" to-layer="5423" to-port="0" />
-		<edge from-layer="5423" from-port="1" to-layer="5424" to-port="1" />
-		<edge from-layer="5424" from-port="2" to-layer="5427" to-port="0" />
-		<edge from-layer="5425" from-port="0" to-layer="5426" to-port="0" />
-		<edge from-layer="5426" from-port="1" to-layer="5427" to-port="1" />
-		<edge from-layer="5427" from-port="2" to-layer="5436" to-port="0" />
-		<edge from-layer="5428" from-port="0" to-layer="5429" to-port="0" />
-		<edge from-layer="5429" from-port="1" to-layer="5432" to-port="0" />
-		<edge from-layer="5430" from-port="0" to-layer="5431" to-port="0" />
-		<edge from-layer="5431" from-port="1" to-layer="5432" to-port="1" />
-		<edge from-layer="5432" from-port="2" to-layer="5434" to-port="0" />
-		<edge from-layer="5433" from-port="0" to-layer="5434" to-port="1" />
-		<edge from-layer="5434" from-port="2" to-layer="5435" to-port="0" />
-		<edge from-layer="5435" from-port="1" to-layer="5436" to-port="1" />
-		<edge from-layer="5436" from-port="2" to-layer="5439" to-port="0" />
-		<edge from-layer="5437" from-port="0" to-layer="5438" to-port="0" />
-		<edge from-layer="5438" from-port="1" to-layer="5439" to-port="1" />
-		<edge from-layer="5439" from-port="2" to-layer="5441" to-port="0" />
-		<edge from-layer="5440" from-port="0" to-layer="5441" to-port="1" />
-		<edge from-layer="5441" from-port="2" to-layer="5569" to-port="0" />
-		<edge from-layer="5441" from-port="2" to-layer="5445" to-port="0" />
-		<edge from-layer="5441" from-port="2" to-layer="5458" to-port="0" />
-		<edge from-layer="5442" from-port="0" to-layer="5445" to-port="1" />
-		<edge from-layer="5443" from-port="0" to-layer="5445" to-port="2" />
-		<edge from-layer="5444" from-port="0" to-layer="5445" to-port="3" />
-		<edge from-layer="5445" from-port="4" to-layer="5447" to-port="0" />
-		<edge from-layer="5446" from-port="0" to-layer="5447" to-port="1" />
-		<edge from-layer="5447" from-port="2" to-layer="5520" to-port="0" />
-		<edge from-layer="5447" from-port="2" to-layer="5451" to-port="0" />
-		<edge from-layer="5448" from-port="0" to-layer="5451" to-port="1" />
-		<edge from-layer="5449" from-port="0" to-layer="5451" to-port="2" />
-		<edge from-layer="5450" from-port="0" to-layer="5451" to-port="3" />
-		<edge from-layer="5451" from-port="4" to-layer="5488" to-port="0" />
-		<edge from-layer="5451" from-port="4" to-layer="5499" to-port="0" />
-		<edge from-layer="5451" from-port="4" to-layer="5505" to-port="0" />
-		<edge from-layer="5451" from-port="4" to-layer="5485" to-port="0" />
-		<edge from-layer="5452" from-port="0" to-layer="5453" to-port="0" />
-		<edge from-layer="5453" from-port="1" to-layer="5481" to-port="0" />
-		<edge from-layer="5454" from-port="0" to-layer="5481" to-port="1" />
-		<edge from-layer="5455" from-port="0" to-layer="5458" to-port="1" />
-		<edge from-layer="5456" from-port="0" to-layer="5458" to-port="2" />
-		<edge from-layer="5457" from-port="0" to-layer="5458" to-port="3" />
-		<edge from-layer="5458" from-port="4" to-layer="5460" to-port="0" />
-		<edge from-layer="5459" from-port="0" to-layer="5460" to-port="1" />
-		<edge from-layer="5460" from-port="2" to-layer="5525" to-port="0" />
-		<edge from-layer="5460" from-port="2" to-layer="5461" to-port="0" />
-		<edge from-layer="5460" from-port="2" to-layer="5553" to-port="0" />
-		<edge from-layer="5461" from-port="1" to-layer="5463" to-port="0" />
-		<edge from-layer="5462" from-port="0" to-layer="5463" to-port="2" />
-		<edge from-layer="5463" from-port="3" to-layer="5477" to-port="0" />
-		<edge from-layer="5464" from-port="0" to-layer="5465" to-port="0" />
-		<edge from-layer="5465" from-port="1" to-layer="5470" to-port="0" />
-		<edge from-layer="5466" from-port="0" to-layer="5469" to-port="1" />
-		<edge from-layer="5467" from-port="0" to-layer="5469" to-port="2" />
-		<edge from-layer="5468" from-port="0" to-layer="5469" to-port="3" />
-		<edge from-layer="5469" from-port="4" to-layer="5470" to-port="1" />
-		<edge from-layer="5470" from-port="2" to-layer="5471" to-port="0" />
-		<edge from-layer="5471" from-port="1" to-layer="5473" to-port="0" />
-		<edge from-layer="5472" from-port="0" to-layer="5473" to-port="2" />
-		<edge from-layer="5473" from-port="3" to-layer="5474" to-port="0" />
-		<edge from-layer="5473" from-port="3" to-layer="5555" to-port="0" />
-		<edge from-layer="5474" from-port="1" to-layer="5476" to-port="0" />
-		<edge from-layer="5475" from-port="0" to-layer="5476" to-port="2" />
-		<edge from-layer="5476" from-port="3" to-layer="5477" to-port="1" />
-		<edge from-layer="5477" from-port="2" to-layer="5479" to-port="0" />
-		<edge from-layer="5478" from-port="0" to-layer="5479" to-port="1" />
-		<edge from-layer="5479" from-port="2" to-layer="5511" to-port="2" />
-		<edge from-layer="5479" from-port="2" to-layer="5481" to-port="2" />
-		<edge from-layer="5480" from-port="0" to-layer="5481" to-port="3" />
-		<edge from-layer="5481" from-port="4" to-layer="5483" to-port="0" />
-		<edge from-layer="5482" from-port="0" to-layer="5483" to-port="2" />
-		<edge from-layer="5483" from-port="3" to-layer="5484" to-port="0" />
-		<edge from-layer="5484" from-port="2" to-layer="5526" to-port="1" />
-		<edge from-layer="5484" from-port="2" to-layer="5485" to-port="1" />
-		<edge from-layer="5485" from-port="2" to-layer="5516" to-port="0" />
-		<edge from-layer="5486" from-port="0" to-layer="5496" to-port="0" />
-		<edge from-layer="5487" from-port="0" to-layer="5496" to-port="1" />
-		<edge from-layer="5488" from-port="1" to-layer="5490" to-port="0" />
-		<edge from-layer="5489" from-port="0" to-layer="5490" to-port="2" />
-		<edge from-layer="5490" from-port="3" to-layer="5491" to-port="0" />
-		<edge from-layer="5491" from-port="2" to-layer="5492" to-port="0" />
-		<edge from-layer="5492" from-port="1" to-layer="5494" to-port="0" />
-		<edge from-layer="5493" from-port="0" to-layer="5494" to-port="1" />
-		<edge from-layer="5494" from-port="2" to-layer="5496" to-port="2" />
-		<edge from-layer="5495" from-port="0" to-layer="5496" to-port="3" />
-		<edge from-layer="5496" from-port="4" to-layer="5499" to-port="1" />
-		<edge from-layer="5496" from-port="4" to-layer="5505" to-port="2" />
-		<edge from-layer="5497" from-port="0" to-layer="5499" to-port="2" />
-		<edge from-layer="5498" from-port="0" to-layer="5499" to-port="3" />
-		<edge from-layer="5499" from-port="4" to-layer="5502" to-port="0" />
-		<edge from-layer="5500" from-port="0" to-layer="5501" to-port="0" />
-		<edge from-layer="5501" from-port="1" to-layer="5502" to-port="1" />
-		<edge from-layer="5502" from-port="2" to-layer="5506" to-port="0" />
-		<edge from-layer="5503" from-port="0" to-layer="5505" to-port="1" />
-		<edge from-layer="5504" from-port="0" to-layer="5505" to-port="3" />
-		<edge from-layer="5505" from-port="4" to-layer="5506" to-port="1" />
-		<edge from-layer="5506" from-port="2" to-layer="5515" to-port="0" />
-		<edge from-layer="5507" from-port="0" to-layer="5508" to-port="0" />
-		<edge from-layer="5508" from-port="1" to-layer="5511" to-port="0" />
-		<edge from-layer="5509" from-port="0" to-layer="5511" to-port="1" />
-		<edge from-layer="5510" from-port="0" to-layer="5511" to-port="3" />
-		<edge from-layer="5511" from-port="4" to-layer="5513" to-port="0" />
-		<edge from-layer="5512" from-port="0" to-layer="5513" to-port="2" />
-		<edge from-layer="5513" from-port="3" to-layer="5514" to-port="0" />
-		<edge from-layer="5514" from-port="2" to-layer="5548" to-port="1" />
-		<edge from-layer="5514" from-port="2" to-layer="5515" to-port="1" />
-		<edge from-layer="5515" from-port="2" to-layer="5516" to-port="1" />
-		<edge from-layer="5516" from-port="2" to-layer="5521" to-port="0" />
-		<edge from-layer="5517" from-port="0" to-layer="5520" to-port="1" />
-		<edge from-layer="5518" from-port="0" to-layer="5520" to-port="2" />
-		<edge from-layer="5519" from-port="0" to-layer="5520" to-port="3" />
-		<edge from-layer="5520" from-port="4" to-layer="5521" to-port="1" />
-		<edge from-layer="5521" from-port="2" to-layer="5579" to-port="0" />
-		<edge from-layer="5521" from-port="2" to-layer="5612" to-port="0" />
-		<edge from-layer="5522" from-port="0" to-layer="5525" to-port="1" />
-		<edge from-layer="5523" from-port="0" to-layer="5525" to-port="2" />
-		<edge from-layer="5524" from-port="0" to-layer="5525" to-port="3" />
-		<edge from-layer="5525" from-port="4" to-layer="5529" to-port="0" />
-		<edge from-layer="5525" from-port="4" to-layer="5526" to-port="0" />
-		<edge from-layer="5525" from-port="4" to-layer="5546" to-port="0" />
-		<edge from-layer="5525" from-port="4" to-layer="5540" to-port="0" />
-		<edge from-layer="5526" from-port="2" to-layer="5549" to-port="0" />
-		<edge from-layer="5527" from-port="0" to-layer="5537" to-port="0" />
-		<edge from-layer="5528" from-port="0" to-layer="5537" to-port="1" />
-		<edge from-layer="5529" from-port="1" to-layer="5531" to-port="0" />
-		<edge from-layer="5530" from-port="0" to-layer="5531" to-port="2" />
-		<edge from-layer="5531" from-port="3" to-layer="5532" to-port="0" />
-		<edge from-layer="5532" from-port="2" to-layer="5533" to-port="0" />
-		<edge from-layer="5533" from-port="1" to-layer="5535" to-port="0" />
-		<edge from-layer="5534" from-port="0" to-layer="5535" to-port="1" />
-		<edge from-layer="5535" from-port="2" to-layer="5537" to-port="2" />
-		<edge from-layer="5536" from-port="0" to-layer="5537" to-port="3" />
-		<edge from-layer="5537" from-port="4" to-layer="5540" to-port="1" />
-		<edge from-layer="5537" from-port="4" to-layer="5546" to-port="2" />
-		<edge from-layer="5538" from-port="0" to-layer="5540" to-port="2" />
-		<edge from-layer="5539" from-port="0" to-layer="5540" to-port="3" />
-		<edge from-layer="5540" from-port="4" to-layer="5543" to-port="0" />
-		<edge from-layer="5541" from-port="0" to-layer="5542" to-port="0" />
-		<edge from-layer="5542" from-port="1" to-layer="5543" to-port="1" />
-		<edge from-layer="5543" from-port="2" to-layer="5547" to-port="0" />
-		<edge from-layer="5544" from-port="0" to-layer="5546" to-port="1" />
-		<edge from-layer="5545" from-port="0" to-layer="5546" to-port="3" />
-		<edge from-layer="5546" from-port="4" to-layer="5547" to-port="1" />
-		<edge from-layer="5547" from-port="2" to-layer="5548" to-port="0" />
-		<edge from-layer="5548" from-port="2" to-layer="5549" to-port="1" />
-		<edge from-layer="5549" from-port="2" to-layer="5554" to-port="0" />
-		<edge from-layer="5550" from-port="0" to-layer="5553" to-port="1" />
-		<edge from-layer="5551" from-port="0" to-layer="5553" to-port="2" />
-		<edge from-layer="5552" from-port="0" to-layer="5553" to-port="3" />
-		<edge from-layer="5553" from-port="4" to-layer="5554" to-port="1" />
-		<edge from-layer="5554" from-port="2" to-layer="5555" to-port="1" />
-		<edge from-layer="5555" from-port="2" to-layer="5612" to-port="1" />
-		<edge from-layer="5555" from-port="2" to-layer="7692" to-port="0" />
-		<edge from-layer="5555" from-port="2" to-layer="5576" to-port="0" />
-		<edge from-layer="5556" from-port="0" to-layer="5557" to-port="0" />
-		<edge from-layer="5557" from-port="1" to-layer="5562" to-port="0" />
-		<edge from-layer="5558" from-port="0" to-layer="5561" to-port="1" />
-		<edge from-layer="5559" from-port="0" to-layer="5561" to-port="2" />
-		<edge from-layer="5560" from-port="0" to-layer="5561" to-port="3" />
-		<edge from-layer="5561" from-port="4" to-layer="5562" to-port="1" />
-		<edge from-layer="5562" from-port="2" to-layer="5563" to-port="0" />
-		<edge from-layer="5563" from-port="1" to-layer="5565" to-port="0" />
-		<edge from-layer="5564" from-port="0" to-layer="5565" to-port="2" />
-		<edge from-layer="5565" from-port="3" to-layer="5572" to-port="0" />
-		<edge from-layer="5566" from-port="0" to-layer="5569" to-port="1" />
-		<edge from-layer="5567" from-port="0" to-layer="5569" to-port="2" />
-		<edge from-layer="5568" from-port="0" to-layer="5569" to-port="3" />
-		<edge from-layer="5569" from-port="4" to-layer="5571" to-port="0" />
-		<edge from-layer="5570" from-port="0" to-layer="5571" to-port="1" />
-		<edge from-layer="5571" from-port="2" to-layer="5572" to-port="1" />
-		<edge from-layer="5572" from-port="2" to-layer="5612" to-port="2" />
-		<edge from-layer="5572" from-port="2" to-layer="7691" to-port="0" />
-		<edge from-layer="5573" from-port="0" to-layer="5591" to-port="0" />
-		<edge from-layer="5574" from-port="0" to-layer="5586" to-port="0" />
-		<edge from-layer="5575" from-port="0" to-layer="5589" to-port="1" />
-		<edge from-layer="5575" from-port="0" to-layer="5586" to-port="1" />
-		<edge from-layer="5576" from-port="1" to-layer="5578" to-port="0" />
-		<edge from-layer="5577" from-port="0" to-layer="5578" to-port="2" />
-		<edge from-layer="5578" from-port="3" to-layer="5588" to-port="0" />
-		<edge from-layer="5578" from-port="3" to-layer="5582" to-port="0" />
-		<edge from-layer="5579" from-port="1" to-layer="5607" to-port="0" />
-		<edge from-layer="5579" from-port="1" to-layer="5581" to-port="0" />
-		<edge from-layer="5580" from-port="0" to-layer="5581" to-port="2" />
-		<edge from-layer="5581" from-port="3" to-layer="5582" to-port="1" />
-		<edge from-layer="5582" from-port="2" to-layer="5584" to-port="0" />
-		<edge from-layer="5583" from-port="0" to-layer="5584" to-port="1" />
-		<edge from-layer="5583" from-port="0" to-layer="5588" to-port="1" />
-		<edge from-layer="5584" from-port="2" to-layer="5586" to-port="2" />
-		<edge from-layer="5585" from-port="0" to-layer="5589" to-port="3" />
-		<edge from-layer="5585" from-port="0" to-layer="5586" to-port="3" />
-		<edge from-layer="5586" from-port="4" to-layer="5591" to-port="1" />
-		<edge from-layer="5587" from-port="0" to-layer="5589" to-port="0" />
-		<edge from-layer="5588" from-port="2" to-layer="5589" to-port="2" />
-		<edge from-layer="5588" from-port="2" to-layer="5596" to-port="2" />
-		<edge from-layer="5589" from-port="4" to-layer="5591" to-port="2" />
-		<edge from-layer="5590" from-port="0" to-layer="5591" to-port="3" />
-		<edge from-layer="5591" from-port="4" to-layer="5598" to-port="0" />
-		<edge from-layer="5592" from-port="0" to-layer="5598" to-port="1" />
-		<edge from-layer="5593" from-port="0" to-layer="5596" to-port="0" />
-		<edge from-layer="5594" from-port="0" to-layer="5596" to-port="1" />
-		<edge from-layer="5595" from-port="0" to-layer="5596" to-port="3" />
-		<edge from-layer="5596" from-port="4" to-layer="5598" to-port="2" />
-		<edge from-layer="5597" from-port="0" to-layer="5598" to-port="3" />
-		<edge from-layer="5598" from-port="4" to-layer="5602" to-port="0" />
-		<edge from-layer="5599" from-port="0" to-layer="5600" to-port="0" />
-		<edge from-layer="5600" from-port="1" to-layer="5602" to-port="1" />
-		<edge from-layer="5601" from-port="0" to-layer="5602" to-port="2" />
-		<edge from-layer="5602" from-port="3" to-layer="5603" to-port="0" />
-		<edge from-layer="5603" from-port="2" to-layer="5604" to-port="0" />
-		<edge from-layer="5603" from-port="2" to-layer="5611" to-port="0" />
-		<edge from-layer="5604" from-port="1" to-layer="5610" to-port="0" />
-		<edge from-layer="5605" from-port="0" to-layer="5607" to-port="1" />
-		<edge from-layer="5606" from-port="0" to-layer="5607" to-port="2" />
-		<edge from-layer="5607" from-port="3" to-layer="5608" to-port="0" />
-		<edge from-layer="5608" from-port="4" to-layer="5609" to-port="0" />
-		<edge from-layer="5609" from-port="1" to-layer="5610" to-port="1" />
-		<edge from-layer="5610" from-port="2" to-layer="5611" to-port="1" />
-		<edge from-layer="5611" from-port="2" to-layer="5612" to-port="3" />
-		<edge from-layer="5612" from-port="4" to-layer="5614" to-port="0" />
-		<edge from-layer="5613" from-port="0" to-layer="5614" to-port="1" />
-		<edge from-layer="5614" from-port="2" to-layer="5616" to-port="0" />
-		<edge from-layer="5615" from-port="0" to-layer="5616" to-port="1" />
-		<edge from-layer="5616" from-port="2" to-layer="5625" to-port="0" />
-		<edge from-layer="5617" from-port="0" to-layer="5618" to-port="0" />
-		<edge from-layer="5618" from-port="1" to-layer="5621" to-port="0" />
-		<edge from-layer="5619" from-port="0" to-layer="5620" to-port="0" />
-		<edge from-layer="5620" from-port="1" to-layer="5621" to-port="1" />
-		<edge from-layer="5621" from-port="2" to-layer="5623" to-port="0" />
-		<edge from-layer="5622" from-port="0" to-layer="5623" to-port="1" />
-		<edge from-layer="5623" from-port="2" to-layer="5624" to-port="0" />
-		<edge from-layer="5624" from-port="1" to-layer="5625" to-port="1" />
-		<edge from-layer="5625" from-port="2" to-layer="5628" to-port="0" />
-		<edge from-layer="5626" from-port="0" to-layer="5627" to-port="0" />
-		<edge from-layer="5627" from-port="1" to-layer="5628" to-port="1" />
-		<edge from-layer="5628" from-port="2" to-layer="5629" to-port="1" />
-		<edge from-layer="5629" from-port="2" to-layer="5630" to-port="0" />
-		<edge from-layer="5630" from-port="2" to-layer="5872" to-port="1" />
-		<edge from-layer="5630" from-port="2" to-layer="5632" to-port="0" />
-		<edge from-layer="5631" from-port="0" to-layer="5632" to-port="1" />
-		<edge from-layer="5632" from-port="2" to-layer="5666" to-port="0" />
-		<edge from-layer="5632" from-port="2" to-layer="5635" to-port="0" />
-		<edge from-layer="5633" from-port="0" to-layer="5634" to-port="0" />
-		<edge from-layer="5634" from-port="1" to-layer="5635" to-port="1" />
-		<edge from-layer="5635" from-port="2" to-layer="5638" to-port="0" />
-		<edge from-layer="5636" from-port="0" to-layer="5637" to-port="0" />
-		<edge from-layer="5637" from-port="1" to-layer="5638" to-port="1" />
-		<edge from-layer="5638" from-port="2" to-layer="5647" to-port="0" />
-		<edge from-layer="5639" from-port="0" to-layer="5640" to-port="0" />
-		<edge from-layer="5640" from-port="1" to-layer="5643" to-port="0" />
-		<edge from-layer="5641" from-port="0" to-layer="5642" to-port="0" />
-		<edge from-layer="5642" from-port="1" to-layer="5643" to-port="1" />
-		<edge from-layer="5643" from-port="2" to-layer="5645" to-port="0" />
-		<edge from-layer="5644" from-port="0" to-layer="5645" to-port="1" />
-		<edge from-layer="5645" from-port="2" to-layer="5646" to-port="0" />
-		<edge from-layer="5646" from-port="1" to-layer="5647" to-port="1" />
-		<edge from-layer="5647" from-port="2" to-layer="5650" to-port="0" />
-		<edge from-layer="5648" from-port="0" to-layer="5649" to-port="0" />
-		<edge from-layer="5649" from-port="1" to-layer="5650" to-port="1" />
-		<edge from-layer="5650" from-port="2" to-layer="5651" to-port="0" />
-		<edge from-layer="5651" from-port="1" to-layer="5660" to-port="0" />
-		<edge from-layer="5652" from-port="0" to-layer="5653" to-port="0" />
-		<edge from-layer="5653" from-port="1" to-layer="5656" to-port="0" />
-		<edge from-layer="5654" from-port="0" to-layer="5655" to-port="0" />
-		<edge from-layer="5655" from-port="1" to-layer="5656" to-port="1" />
-		<edge from-layer="5656" from-port="2" to-layer="5658" to-port="0" />
-		<edge from-layer="5657" from-port="0" to-layer="5658" to-port="1" />
-		<edge from-layer="5658" from-port="2" to-layer="5659" to-port="0" />
-		<edge from-layer="5659" from-port="1" to-layer="5660" to-port="1" />
-		<edge from-layer="5660" from-port="2" to-layer="5663" to-port="0" />
-		<edge from-layer="5661" from-port="0" to-layer="5662" to-port="0" />
-		<edge from-layer="5662" from-port="1" to-layer="5663" to-port="1" />
-		<edge from-layer="5663" from-port="2" to-layer="5871" to-port="0" />
-		<edge from-layer="5664" from-port="0" to-layer="5665" to-port="0" />
-		<edge from-layer="5665" from-port="1" to-layer="5666" to-port="1" />
-		<edge from-layer="5666" from-port="2" to-layer="5669" to-port="0" />
-		<edge from-layer="5667" from-port="0" to-layer="5668" to-port="0" />
-		<edge from-layer="5668" from-port="1" to-layer="5669" to-port="1" />
-		<edge from-layer="5669" from-port="2" to-layer="5678" to-port="0" />
-		<edge from-layer="5670" from-port="0" to-layer="5671" to-port="0" />
-		<edge from-layer="5671" from-port="1" to-layer="5674" to-port="0" />
-		<edge from-layer="5672" from-port="0" to-layer="5673" to-port="0" />
-		<edge from-layer="5673" from-port="1" to-layer="5674" to-port="1" />
-		<edge from-layer="5674" from-port="2" to-layer="5676" to-port="0" />
-		<edge from-layer="5675" from-port="0" to-layer="5676" to-port="1" />
-		<edge from-layer="5676" from-port="2" to-layer="5677" to-port="0" />
-		<edge from-layer="5677" from-port="1" to-layer="5678" to-port="1" />
-		<edge from-layer="5678" from-port="2" to-layer="5681" to-port="0" />
-		<edge from-layer="5679" from-port="0" to-layer="5680" to-port="0" />
-		<edge from-layer="5680" from-port="1" to-layer="5681" to-port="1" />
-		<edge from-layer="5681" from-port="2" to-layer="5683" to-port="0" />
-		<edge from-layer="5682" from-port="0" to-layer="5683" to-port="1" />
-		<edge from-layer="5683" from-port="2" to-layer="5811" to-port="0" />
-		<edge from-layer="5683" from-port="2" to-layer="5687" to-port="0" />
-		<edge from-layer="5683" from-port="2" to-layer="5700" to-port="0" />
-		<edge from-layer="5684" from-port="0" to-layer="5687" to-port="1" />
-		<edge from-layer="5685" from-port="0" to-layer="5687" to-port="2" />
-		<edge from-layer="5686" from-port="0" to-layer="5687" to-port="3" />
-		<edge from-layer="5687" from-port="4" to-layer="5689" to-port="0" />
-		<edge from-layer="5688" from-port="0" to-layer="5689" to-port="1" />
-		<edge from-layer="5689" from-port="2" to-layer="5693" to-port="0" />
-		<edge from-layer="5689" from-port="2" to-layer="5762" to-port="0" />
-		<edge from-layer="5690" from-port="0" to-layer="5693" to-port="1" />
-		<edge from-layer="5691" from-port="0" to-layer="5693" to-port="2" />
-		<edge from-layer="5692" from-port="0" to-layer="5693" to-port="3" />
-		<edge from-layer="5693" from-port="4" to-layer="5747" to-port="0" />
-		<edge from-layer="5693" from-port="4" to-layer="5741" to-port="0" />
-		<edge from-layer="5693" from-port="4" to-layer="5730" to-port="0" />
-		<edge from-layer="5693" from-port="4" to-layer="5727" to-port="0" />
-		<edge from-layer="5694" from-port="0" to-layer="5695" to-port="0" />
-		<edge from-layer="5695" from-port="1" to-layer="5723" to-port="0" />
-		<edge from-layer="5696" from-port="0" to-layer="5723" to-port="1" />
-		<edge from-layer="5697" from-port="0" to-layer="5700" to-port="1" />
-		<edge from-layer="5698" from-port="0" to-layer="5700" to-port="2" />
-		<edge from-layer="5699" from-port="0" to-layer="5700" to-port="3" />
-		<edge from-layer="5700" from-port="4" to-layer="5702" to-port="0" />
-		<edge from-layer="5701" from-port="0" to-layer="5702" to-port="1" />
-		<edge from-layer="5702" from-port="2" to-layer="5795" to-port="0" />
-		<edge from-layer="5702" from-port="2" to-layer="5767" to-port="0" />
-		<edge from-layer="5702" from-port="2" to-layer="5703" to-port="0" />
-		<edge from-layer="5703" from-port="1" to-layer="5705" to-port="0" />
-		<edge from-layer="5704" from-port="0" to-layer="5705" to-port="2" />
-		<edge from-layer="5705" from-port="3" to-layer="5719" to-port="0" />
-		<edge from-layer="5706" from-port="0" to-layer="5707" to-port="0" />
-		<edge from-layer="5707" from-port="1" to-layer="5712" to-port="0" />
-		<edge from-layer="5708" from-port="0" to-layer="5711" to-port="1" />
-		<edge from-layer="5709" from-port="0" to-layer="5711" to-port="2" />
-		<edge from-layer="5710" from-port="0" to-layer="5711" to-port="3" />
-		<edge from-layer="5711" from-port="4" to-layer="5712" to-port="1" />
-		<edge from-layer="5712" from-port="2" to-layer="5713" to-port="0" />
-		<edge from-layer="5713" from-port="1" to-layer="5715" to-port="0" />
-		<edge from-layer="5714" from-port="0" to-layer="5715" to-port="2" />
-		<edge from-layer="5715" from-port="3" to-layer="5797" to-port="0" />
-		<edge from-layer="5715" from-port="3" to-layer="5716" to-port="0" />
-		<edge from-layer="5716" from-port="1" to-layer="5718" to-port="0" />
-		<edge from-layer="5717" from-port="0" to-layer="5718" to-port="2" />
-		<edge from-layer="5718" from-port="3" to-layer="5719" to-port="1" />
-		<edge from-layer="5719" from-port="2" to-layer="5721" to-port="0" />
-		<edge from-layer="5720" from-port="0" to-layer="5721" to-port="1" />
-		<edge from-layer="5721" from-port="2" to-layer="5753" to-port="2" />
-		<edge from-layer="5721" from-port="2" to-layer="5723" to-port="2" />
-		<edge from-layer="5722" from-port="0" to-layer="5723" to-port="3" />
-		<edge from-layer="5723" from-port="4" to-layer="5725" to-port="0" />
-		<edge from-layer="5724" from-port="0" to-layer="5725" to-port="2" />
-		<edge from-layer="5725" from-port="3" to-layer="5726" to-port="0" />
-		<edge from-layer="5726" from-port="2" to-layer="5768" to-port="1" />
-		<edge from-layer="5726" from-port="2" to-layer="5727" to-port="1" />
-		<edge from-layer="5727" from-port="2" to-layer="5758" to-port="0" />
-		<edge from-layer="5728" from-port="0" to-layer="5738" to-port="0" />
-		<edge from-layer="5729" from-port="0" to-layer="5738" to-port="1" />
-		<edge from-layer="5730" from-port="1" to-layer="5732" to-port="0" />
-		<edge from-layer="5731" from-port="0" to-layer="5732" to-port="2" />
-		<edge from-layer="5732" from-port="3" to-layer="5733" to-port="0" />
-		<edge from-layer="5733" from-port="2" to-layer="5734" to-port="0" />
-		<edge from-layer="5734" from-port="1" to-layer="5736" to-port="0" />
-		<edge from-layer="5735" from-port="0" to-layer="5736" to-port="1" />
-		<edge from-layer="5736" from-port="2" to-layer="5738" to-port="2" />
-		<edge from-layer="5737" from-port="0" to-layer="5738" to-port="3" />
-		<edge from-layer="5738" from-port="4" to-layer="5741" to-port="1" />
-		<edge from-layer="5738" from-port="4" to-layer="5747" to-port="2" />
-		<edge from-layer="5739" from-port="0" to-layer="5741" to-port="2" />
-		<edge from-layer="5740" from-port="0" to-layer="5741" to-port="3" />
-		<edge from-layer="5741" from-port="4" to-layer="5744" to-port="0" />
-		<edge from-layer="5742" from-port="0" to-layer="5743" to-port="0" />
-		<edge from-layer="5743" from-port="1" to-layer="5744" to-port="1" />
-		<edge from-layer="5744" from-port="2" to-layer="5748" to-port="0" />
-		<edge from-layer="5745" from-port="0" to-layer="5747" to-port="1" />
-		<edge from-layer="5746" from-port="0" to-layer="5747" to-port="3" />
-		<edge from-layer="5747" from-port="4" to-layer="5748" to-port="1" />
-		<edge from-layer="5748" from-port="2" to-layer="5757" to-port="0" />
-		<edge from-layer="5749" from-port="0" to-layer="5750" to-port="0" />
-		<edge from-layer="5750" from-port="1" to-layer="5753" to-port="0" />
-		<edge from-layer="5751" from-port="0" to-layer="5753" to-port="1" />
-		<edge from-layer="5752" from-port="0" to-layer="5753" to-port="3" />
-		<edge from-layer="5753" from-port="4" to-layer="5755" to-port="0" />
-		<edge from-layer="5754" from-port="0" to-layer="5755" to-port="2" />
-		<edge from-layer="5755" from-port="3" to-layer="5756" to-port="0" />
-		<edge from-layer="5756" from-port="2" to-layer="5757" to-port="1" />
-		<edge from-layer="5756" from-port="2" to-layer="5790" to-port="1" />
-		<edge from-layer="5757" from-port="2" to-layer="5758" to-port="1" />
-		<edge from-layer="5758" from-port="2" to-layer="5763" to-port="0" />
-		<edge from-layer="5759" from-port="0" to-layer="5762" to-port="1" />
-		<edge from-layer="5760" from-port="0" to-layer="5762" to-port="2" />
-		<edge from-layer="5761" from-port="0" to-layer="5762" to-port="3" />
-		<edge from-layer="5762" from-port="4" to-layer="5763" to-port="1" />
-		<edge from-layer="5763" from-port="2" to-layer="5821" to-port="0" />
-		<edge from-layer="5763" from-port="2" to-layer="5854" to-port="0" />
-		<edge from-layer="5764" from-port="0" to-layer="5767" to-port="1" />
-		<edge from-layer="5765" from-port="0" to-layer="5767" to-port="2" />
-		<edge from-layer="5766" from-port="0" to-layer="5767" to-port="3" />
-		<edge from-layer="5767" from-port="4" to-layer="5782" to-port="0" />
-		<edge from-layer="5767" from-port="4" to-layer="5788" to-port="0" />
-		<edge from-layer="5767" from-port="4" to-layer="5771" to-port="0" />
-		<edge from-layer="5767" from-port="4" to-layer="5768" to-port="0" />
-		<edge from-layer="5768" from-port="2" to-layer="5791" to-port="0" />
-		<edge from-layer="5769" from-port="0" to-layer="5779" to-port="0" />
-		<edge from-layer="5770" from-port="0" to-layer="5779" to-port="1" />
-		<edge from-layer="5771" from-port="1" to-layer="5773" to-port="0" />
-		<edge from-layer="5772" from-port="0" to-layer="5773" to-port="2" />
-		<edge from-layer="5773" from-port="3" to-layer="5774" to-port="0" />
-		<edge from-layer="5774" from-port="2" to-layer="5775" to-port="0" />
-		<edge from-layer="5775" from-port="1" to-layer="5777" to-port="0" />
-		<edge from-layer="5776" from-port="0" to-layer="5777" to-port="1" />
-		<edge from-layer="5777" from-port="2" to-layer="5779" to-port="2" />
-		<edge from-layer="5778" from-port="0" to-layer="5779" to-port="3" />
-		<edge from-layer="5779" from-port="4" to-layer="5788" to-port="2" />
-		<edge from-layer="5779" from-port="4" to-layer="5782" to-port="1" />
-		<edge from-layer="5780" from-port="0" to-layer="5782" to-port="2" />
-		<edge from-layer="5781" from-port="0" to-layer="5782" to-port="3" />
-		<edge from-layer="5782" from-port="4" to-layer="5785" to-port="0" />
-		<edge from-layer="5783" from-port="0" to-layer="5784" to-port="0" />
-		<edge from-layer="5784" from-port="1" to-layer="5785" to-port="1" />
-		<edge from-layer="5785" from-port="2" to-layer="5789" to-port="0" />
-		<edge from-layer="5786" from-port="0" to-layer="5788" to-port="1" />
-		<edge from-layer="5787" from-port="0" to-layer="5788" to-port="3" />
-		<edge from-layer="5788" from-port="4" to-layer="5789" to-port="1" />
-		<edge from-layer="5789" from-port="2" to-layer="5790" to-port="0" />
-		<edge from-layer="5790" from-port="2" to-layer="5791" to-port="1" />
-		<edge from-layer="5791" from-port="2" to-layer="5796" to-port="0" />
-		<edge from-layer="5792" from-port="0" to-layer="5795" to-port="1" />
-		<edge from-layer="5793" from-port="0" to-layer="5795" to-port="2" />
-		<edge from-layer="5794" from-port="0" to-layer="5795" to-port="3" />
-		<edge from-layer="5795" from-port="4" to-layer="5796" to-port="1" />
-		<edge from-layer="5796" from-port="2" to-layer="5797" to-port="1" />
-		<edge from-layer="5797" from-port="2" to-layer="5818" to-port="0" />
-		<edge from-layer="5797" from-port="2" to-layer="5854" to-port="1" />
-		<edge from-layer="5797" from-port="2" to-layer="7690" to-port="0" />
-		<edge from-layer="5798" from-port="0" to-layer="5799" to-port="0" />
-		<edge from-layer="5799" from-port="1" to-layer="5804" to-port="0" />
-		<edge from-layer="5800" from-port="0" to-layer="5803" to-port="1" />
-		<edge from-layer="5801" from-port="0" to-layer="5803" to-port="2" />
-		<edge from-layer="5802" from-port="0" to-layer="5803" to-port="3" />
-		<edge from-layer="5803" from-port="4" to-layer="5804" to-port="1" />
-		<edge from-layer="5804" from-port="2" to-layer="5805" to-port="0" />
-		<edge from-layer="5805" from-port="1" to-layer="5807" to-port="0" />
-		<edge from-layer="5806" from-port="0" to-layer="5807" to-port="2" />
-		<edge from-layer="5807" from-port="3" to-layer="5814" to-port="0" />
-		<edge from-layer="5808" from-port="0" to-layer="5811" to-port="1" />
-		<edge from-layer="5809" from-port="0" to-layer="5811" to-port="2" />
-		<edge from-layer="5810" from-port="0" to-layer="5811" to-port="3" />
-		<edge from-layer="5811" from-port="4" to-layer="5813" to-port="0" />
-		<edge from-layer="5812" from-port="0" to-layer="5813" to-port="1" />
-		<edge from-layer="5813" from-port="2" to-layer="5814" to-port="1" />
-		<edge from-layer="5814" from-port="2" to-layer="7689" to-port="0" />
-		<edge from-layer="5814" from-port="2" to-layer="5854" to-port="2" />
-		<edge from-layer="5815" from-port="0" to-layer="5833" to-port="0" />
-		<edge from-layer="5816" from-port="0" to-layer="5828" to-port="0" />
-		<edge from-layer="5817" from-port="0" to-layer="5828" to-port="1" />
-		<edge from-layer="5817" from-port="0" to-layer="5831" to-port="1" />
-		<edge from-layer="5818" from-port="1" to-layer="5820" to-port="0" />
-		<edge from-layer="5819" from-port="0" to-layer="5820" to-port="2" />
-		<edge from-layer="5820" from-port="3" to-layer="5824" to-port="0" />
-		<edge from-layer="5820" from-port="3" to-layer="5830" to-port="0" />
-		<edge from-layer="5821" from-port="1" to-layer="5849" to-port="0" />
-		<edge from-layer="5821" from-port="1" to-layer="5823" to-port="0" />
-		<edge from-layer="5822" from-port="0" to-layer="5823" to-port="2" />
-		<edge from-layer="5823" from-port="3" to-layer="5824" to-port="1" />
-		<edge from-layer="5824" from-port="2" to-layer="5826" to-port="0" />
-		<edge from-layer="5825" from-port="0" to-layer="5826" to-port="1" />
-		<edge from-layer="5825" from-port="0" to-layer="5830" to-port="1" />
-		<edge from-layer="5826" from-port="2" to-layer="5828" to-port="2" />
-		<edge from-layer="5827" from-port="0" to-layer="5831" to-port="3" />
-		<edge from-layer="5827" from-port="0" to-layer="5828" to-port="3" />
-		<edge from-layer="5828" from-port="4" to-layer="5833" to-port="1" />
-		<edge from-layer="5829" from-port="0" to-layer="5831" to-port="0" />
-		<edge from-layer="5830" from-port="2" to-layer="5838" to-port="2" />
-		<edge from-layer="5830" from-port="2" to-layer="5831" to-port="2" />
-		<edge from-layer="5831" from-port="4" to-layer="5833" to-port="2" />
-		<edge from-layer="5832" from-port="0" to-layer="5833" to-port="3" />
-		<edge from-layer="5833" from-port="4" to-layer="5840" to-port="0" />
-		<edge from-layer="5834" from-port="0" to-layer="5840" to-port="1" />
-		<edge from-layer="5835" from-port="0" to-layer="5838" to-port="0" />
-		<edge from-layer="5836" from-port="0" to-layer="5838" to-port="1" />
-		<edge from-layer="5837" from-port="0" to-layer="5838" to-port="3" />
-		<edge from-layer="5838" from-port="4" to-layer="5840" to-port="2" />
-		<edge from-layer="5839" from-port="0" to-layer="5840" to-port="3" />
-		<edge from-layer="5840" from-port="4" to-layer="5844" to-port="0" />
-		<edge from-layer="5841" from-port="0" to-layer="5842" to-port="0" />
-		<edge from-layer="5842" from-port="1" to-layer="5844" to-port="1" />
-		<edge from-layer="5843" from-port="0" to-layer="5844" to-port="2" />
-		<edge from-layer="5844" from-port="3" to-layer="5845" to-port="0" />
-		<edge from-layer="5845" from-port="2" to-layer="5846" to-port="0" />
-		<edge from-layer="5845" from-port="2" to-layer="5853" to-port="0" />
-		<edge from-layer="5846" from-port="1" to-layer="5852" to-port="0" />
-		<edge from-layer="5847" from-port="0" to-layer="5849" to-port="1" />
-		<edge from-layer="5848" from-port="0" to-layer="5849" to-port="2" />
-		<edge from-layer="5849" from-port="3" to-layer="5850" to-port="0" />
-		<edge from-layer="5850" from-port="4" to-layer="5851" to-port="0" />
-		<edge from-layer="5851" from-port="1" to-layer="5852" to-port="1" />
-		<edge from-layer="5852" from-port="2" to-layer="5853" to-port="1" />
-		<edge from-layer="5853" from-port="2" to-layer="5854" to-port="3" />
-		<edge from-layer="5854" from-port="4" to-layer="5856" to-port="0" />
-		<edge from-layer="5855" from-port="0" to-layer="5856" to-port="1" />
-		<edge from-layer="5856" from-port="2" to-layer="5858" to-port="0" />
-		<edge from-layer="5857" from-port="0" to-layer="5858" to-port="1" />
-		<edge from-layer="5858" from-port="2" to-layer="5867" to-port="0" />
-		<edge from-layer="5859" from-port="0" to-layer="5860" to-port="0" />
-		<edge from-layer="5860" from-port="1" to-layer="5863" to-port="0" />
-		<edge from-layer="5861" from-port="0" to-layer="5862" to-port="0" />
-		<edge from-layer="5862" from-port="1" to-layer="5863" to-port="1" />
-		<edge from-layer="5863" from-port="2" to-layer="5865" to-port="0" />
-		<edge from-layer="5864" from-port="0" to-layer="5865" to-port="1" />
-		<edge from-layer="5865" from-port="2" to-layer="5866" to-port="0" />
-		<edge from-layer="5866" from-port="1" to-layer="5867" to-port="1" />
-		<edge from-layer="5867" from-port="2" to-layer="5870" to-port="0" />
-		<edge from-layer="5868" from-port="0" to-layer="5869" to-port="0" />
-		<edge from-layer="5869" from-port="1" to-layer="5870" to-port="1" />
-		<edge from-layer="5870" from-port="2" to-layer="5871" to-port="1" />
-		<edge from-layer="5871" from-port="2" to-layer="5872" to-port="0" />
-		<edge from-layer="5872" from-port="2" to-layer="6114" to-port="1" />
-		<edge from-layer="5872" from-port="2" to-layer="5874" to-port="0" />
-		<edge from-layer="5873" from-port="0" to-layer="5874" to-port="1" />
-		<edge from-layer="5874" from-port="2" to-layer="5877" to-port="0" />
-		<edge from-layer="5874" from-port="2" to-layer="5908" to-port="0" />
-		<edge from-layer="5875" from-port="0" to-layer="5876" to-port="0" />
-		<edge from-layer="5876" from-port="1" to-layer="5877" to-port="1" />
-		<edge from-layer="5877" from-port="2" to-layer="5880" to-port="0" />
-		<edge from-layer="5878" from-port="0" to-layer="5879" to-port="0" />
-		<edge from-layer="5879" from-port="1" to-layer="5880" to-port="1" />
-		<edge from-layer="5880" from-port="2" to-layer="5889" to-port="0" />
-		<edge from-layer="5881" from-port="0" to-layer="5882" to-port="0" />
-		<edge from-layer="5882" from-port="1" to-layer="5885" to-port="0" />
-		<edge from-layer="5883" from-port="0" to-layer="5884" to-port="0" />
-		<edge from-layer="5884" from-port="1" to-layer="5885" to-port="1" />
-		<edge from-layer="5885" from-port="2" to-layer="5887" to-port="0" />
-		<edge from-layer="5886" from-port="0" to-layer="5887" to-port="1" />
-		<edge from-layer="5887" from-port="2" to-layer="5888" to-port="0" />
-		<edge from-layer="5888" from-port="1" to-layer="5889" to-port="1" />
-		<edge from-layer="5889" from-port="2" to-layer="5892" to-port="0" />
-		<edge from-layer="5890" from-port="0" to-layer="5891" to-port="0" />
-		<edge from-layer="5891" from-port="1" to-layer="5892" to-port="1" />
-		<edge from-layer="5892" from-port="2" to-layer="5893" to-port="0" />
-		<edge from-layer="5893" from-port="1" to-layer="5902" to-port="0" />
-		<edge from-layer="5894" from-port="0" to-layer="5895" to-port="0" />
-		<edge from-layer="5895" from-port="1" to-layer="5898" to-port="0" />
-		<edge from-layer="5896" from-port="0" to-layer="5897" to-port="0" />
-		<edge from-layer="5897" from-port="1" to-layer="5898" to-port="1" />
-		<edge from-layer="5898" from-port="2" to-layer="5900" to-port="0" />
-		<edge from-layer="5899" from-port="0" to-layer="5900" to-port="1" />
-		<edge from-layer="5900" from-port="2" to-layer="5901" to-port="0" />
-		<edge from-layer="5901" from-port="1" to-layer="5902" to-port="1" />
-		<edge from-layer="5902" from-port="2" to-layer="5905" to-port="0" />
-		<edge from-layer="5903" from-port="0" to-layer="5904" to-port="0" />
-		<edge from-layer="5904" from-port="1" to-layer="5905" to-port="1" />
-		<edge from-layer="5905" from-port="2" to-layer="6113" to-port="0" />
-		<edge from-layer="5906" from-port="0" to-layer="5907" to-port="0" />
-		<edge from-layer="5907" from-port="1" to-layer="5908" to-port="1" />
-		<edge from-layer="5908" from-port="2" to-layer="5911" to-port="0" />
-		<edge from-layer="5909" from-port="0" to-layer="5910" to-port="0" />
-		<edge from-layer="5910" from-port="1" to-layer="5911" to-port="1" />
-		<edge from-layer="5911" from-port="2" to-layer="5920" to-port="0" />
-		<edge from-layer="5912" from-port="0" to-layer="5913" to-port="0" />
-		<edge from-layer="5913" from-port="1" to-layer="5916" to-port="0" />
-		<edge from-layer="5914" from-port="0" to-layer="5915" to-port="0" />
-		<edge from-layer="5915" from-port="1" to-layer="5916" to-port="1" />
-		<edge from-layer="5916" from-port="2" to-layer="5918" to-port="0" />
-		<edge from-layer="5917" from-port="0" to-layer="5918" to-port="1" />
-		<edge from-layer="5918" from-port="2" to-layer="5919" to-port="0" />
-		<edge from-layer="5919" from-port="1" to-layer="5920" to-port="1" />
-		<edge from-layer="5920" from-port="2" to-layer="5923" to-port="0" />
-		<edge from-layer="5921" from-port="0" to-layer="5922" to-port="0" />
-		<edge from-layer="5922" from-port="1" to-layer="5923" to-port="1" />
-		<edge from-layer="5923" from-port="2" to-layer="5925" to-port="0" />
-		<edge from-layer="5924" from-port="0" to-layer="5925" to-port="1" />
-		<edge from-layer="5925" from-port="2" to-layer="5929" to-port="0" />
-		<edge from-layer="5925" from-port="2" to-layer="6053" to-port="0" />
-		<edge from-layer="5925" from-port="2" to-layer="5942" to-port="0" />
-		<edge from-layer="5926" from-port="0" to-layer="5929" to-port="1" />
-		<edge from-layer="5927" from-port="0" to-layer="5929" to-port="2" />
-		<edge from-layer="5928" from-port="0" to-layer="5929" to-port="3" />
-		<edge from-layer="5929" from-port="4" to-layer="5931" to-port="0" />
-		<edge from-layer="5930" from-port="0" to-layer="5931" to-port="1" />
-		<edge from-layer="5931" from-port="2" to-layer="6004" to-port="0" />
-		<edge from-layer="5931" from-port="2" to-layer="5935" to-port="0" />
-		<edge from-layer="5932" from-port="0" to-layer="5935" to-port="1" />
-		<edge from-layer="5933" from-port="0" to-layer="5935" to-port="2" />
-		<edge from-layer="5934" from-port="0" to-layer="5935" to-port="3" />
-		<edge from-layer="5935" from-port="4" to-layer="5969" to-port="0" />
-		<edge from-layer="5935" from-port="4" to-layer="5972" to-port="0" />
-		<edge from-layer="5935" from-port="4" to-layer="5983" to-port="0" />
-		<edge from-layer="5935" from-port="4" to-layer="5989" to-port="0" />
-		<edge from-layer="5936" from-port="0" to-layer="5937" to-port="0" />
-		<edge from-layer="5937" from-port="1" to-layer="5965" to-port="0" />
-		<edge from-layer="5938" from-port="0" to-layer="5965" to-port="1" />
-		<edge from-layer="5939" from-port="0" to-layer="5942" to-port="1" />
-		<edge from-layer="5940" from-port="0" to-layer="5942" to-port="2" />
-		<edge from-layer="5941" from-port="0" to-layer="5942" to-port="3" />
-		<edge from-layer="5942" from-port="4" to-layer="5944" to-port="0" />
-		<edge from-layer="5943" from-port="0" to-layer="5944" to-port="1" />
-		<edge from-layer="5944" from-port="2" to-layer="6037" to-port="0" />
-		<edge from-layer="5944" from-port="2" to-layer="6009" to-port="0" />
-		<edge from-layer="5944" from-port="2" to-layer="5945" to-port="0" />
-		<edge from-layer="5945" from-port="1" to-layer="5947" to-port="0" />
-		<edge from-layer="5946" from-port="0" to-layer="5947" to-port="2" />
-		<edge from-layer="5947" from-port="3" to-layer="5961" to-port="0" />
-		<edge from-layer="5948" from-port="0" to-layer="5949" to-port="0" />
-		<edge from-layer="5949" from-port="1" to-layer="5954" to-port="0" />
-		<edge from-layer="5950" from-port="0" to-layer="5953" to-port="1" />
-		<edge from-layer="5951" from-port="0" to-layer="5953" to-port="2" />
-		<edge from-layer="5952" from-port="0" to-layer="5953" to-port="3" />
-		<edge from-layer="5953" from-port="4" to-layer="5954" to-port="1" />
-		<edge from-layer="5954" from-port="2" to-layer="5955" to-port="0" />
-		<edge from-layer="5955" from-port="1" to-layer="5957" to-port="0" />
-		<edge from-layer="5956" from-port="0" to-layer="5957" to-port="2" />
-		<edge from-layer="5957" from-port="3" to-layer="5958" to-port="0" />
-		<edge from-layer="5957" from-port="3" to-layer="6039" to-port="0" />
-		<edge from-layer="5958" from-port="1" to-layer="5960" to-port="0" />
-		<edge from-layer="5959" from-port="0" to-layer="5960" to-port="2" />
-		<edge from-layer="5960" from-port="3" to-layer="5961" to-port="1" />
-		<edge from-layer="5961" from-port="2" to-layer="5963" to-port="0" />
-		<edge from-layer="5962" from-port="0" to-layer="5963" to-port="1" />
-		<edge from-layer="5963" from-port="2" to-layer="5995" to-port="2" />
-		<edge from-layer="5963" from-port="2" to-layer="5965" to-port="2" />
-		<edge from-layer="5964" from-port="0" to-layer="5965" to-port="3" />
-		<edge from-layer="5965" from-port="4" to-layer="5967" to-port="0" />
-		<edge from-layer="5966" from-port="0" to-layer="5967" to-port="2" />
-		<edge from-layer="5967" from-port="3" to-layer="5968" to-port="0" />
-		<edge from-layer="5968" from-port="2" to-layer="5969" to-port="1" />
-		<edge from-layer="5968" from-port="2" to-layer="6010" to-port="1" />
-		<edge from-layer="5969" from-port="2" to-layer="6000" to-port="0" />
-		<edge from-layer="5970" from-port="0" to-layer="5980" to-port="0" />
-		<edge from-layer="5971" from-port="0" to-layer="5980" to-port="1" />
-		<edge from-layer="5972" from-port="1" to-layer="5974" to-port="0" />
-		<edge from-layer="5973" from-port="0" to-layer="5974" to-port="2" />
-		<edge from-layer="5974" from-port="3" to-layer="5975" to-port="0" />
-		<edge from-layer="5975" from-port="2" to-layer="5976" to-port="0" />
-		<edge from-layer="5976" from-port="1" to-layer="5978" to-port="0" />
-		<edge from-layer="5977" from-port="0" to-layer="5978" to-port="1" />
-		<edge from-layer="5978" from-port="2" to-layer="5980" to-port="2" />
-		<edge from-layer="5979" from-port="0" to-layer="5980" to-port="3" />
-		<edge from-layer="5980" from-port="4" to-layer="5983" to-port="1" />
-		<edge from-layer="5980" from-port="4" to-layer="5989" to-port="2" />
-		<edge from-layer="5981" from-port="0" to-layer="5983" to-port="2" />
-		<edge from-layer="5982" from-port="0" to-layer="5983" to-port="3" />
-		<edge from-layer="5983" from-port="4" to-layer="5986" to-port="0" />
-		<edge from-layer="5984" from-port="0" to-layer="5985" to-port="0" />
-		<edge from-layer="5985" from-port="1" to-layer="5986" to-port="1" />
-		<edge from-layer="5986" from-port="2" to-layer="5990" to-port="0" />
-		<edge from-layer="5987" from-port="0" to-layer="5989" to-port="1" />
-		<edge from-layer="5988" from-port="0" to-layer="5989" to-port="3" />
-		<edge from-layer="5989" from-port="4" to-layer="5990" to-port="1" />
-		<edge from-layer="5990" from-port="2" to-layer="5999" to-port="0" />
-		<edge from-layer="5991" from-port="0" to-layer="5992" to-port="0" />
-		<edge from-layer="5992" from-port="1" to-layer="5995" to-port="0" />
-		<edge from-layer="5993" from-port="0" to-layer="5995" to-port="1" />
-		<edge from-layer="5994" from-port="0" to-layer="5995" to-port="3" />
-		<edge from-layer="5995" from-port="4" to-layer="5997" to-port="0" />
-		<edge from-layer="5996" from-port="0" to-layer="5997" to-port="2" />
-		<edge from-layer="5997" from-port="3" to-layer="5998" to-port="0" />
-		<edge from-layer="5998" from-port="2" to-layer="6032" to-port="1" />
-		<edge from-layer="5998" from-port="2" to-layer="5999" to-port="1" />
-		<edge from-layer="5999" from-port="2" to-layer="6000" to-port="1" />
-		<edge from-layer="6000" from-port="2" to-layer="6005" to-port="0" />
-		<edge from-layer="6001" from-port="0" to-layer="6004" to-port="1" />
-		<edge from-layer="6002" from-port="0" to-layer="6004" to-port="2" />
-		<edge from-layer="6003" from-port="0" to-layer="6004" to-port="3" />
-		<edge from-layer="6004" from-port="4" to-layer="6005" to-port="1" />
-		<edge from-layer="6005" from-port="2" to-layer="6096" to-port="0" />
-		<edge from-layer="6005" from-port="2" to-layer="6063" to-port="0" />
-		<edge from-layer="6006" from-port="0" to-layer="6009" to-port="1" />
-		<edge from-layer="6007" from-port="0" to-layer="6009" to-port="2" />
-		<edge from-layer="6008" from-port="0" to-layer="6009" to-port="3" />
-		<edge from-layer="6009" from-port="4" to-layer="6013" to-port="0" />
-		<edge from-layer="6009" from-port="4" to-layer="6010" to-port="0" />
-		<edge from-layer="6009" from-port="4" to-layer="6030" to-port="0" />
-		<edge from-layer="6009" from-port="4" to-layer="6024" to-port="0" />
-		<edge from-layer="6010" from-port="2" to-layer="6033" to-port="0" />
-		<edge from-layer="6011" from-port="0" to-layer="6021" to-port="0" />
-		<edge from-layer="6012" from-port="0" to-layer="6021" to-port="1" />
-		<edge from-layer="6013" from-port="1" to-layer="6015" to-port="0" />
-		<edge from-layer="6014" from-port="0" to-layer="6015" to-port="2" />
-		<edge from-layer="6015" from-port="3" to-layer="6016" to-port="0" />
-		<edge from-layer="6016" from-port="2" to-layer="6017" to-port="0" />
-		<edge from-layer="6017" from-port="1" to-layer="6019" to-port="0" />
-		<edge from-layer="6018" from-port="0" to-layer="6019" to-port="1" />
-		<edge from-layer="6019" from-port="2" to-layer="6021" to-port="2" />
-		<edge from-layer="6020" from-port="0" to-layer="6021" to-port="3" />
-		<edge from-layer="6021" from-port="4" to-layer="6024" to-port="1" />
-		<edge from-layer="6021" from-port="4" to-layer="6030" to-port="2" />
-		<edge from-layer="6022" from-port="0" to-layer="6024" to-port="2" />
-		<edge from-layer="6023" from-port="0" to-layer="6024" to-port="3" />
-		<edge from-layer="6024" from-port="4" to-layer="6027" to-port="0" />
-		<edge from-layer="6025" from-port="0" to-layer="6026" to-port="0" />
-		<edge from-layer="6026" from-port="1" to-layer="6027" to-port="1" />
-		<edge from-layer="6027" from-port="2" to-layer="6031" to-port="0" />
-		<edge from-layer="6028" from-port="0" to-layer="6030" to-port="1" />
-		<edge from-layer="6029" from-port="0" to-layer="6030" to-port="3" />
-		<edge from-layer="6030" from-port="4" to-layer="6031" to-port="1" />
-		<edge from-layer="6031" from-port="2" to-layer="6032" to-port="0" />
-		<edge from-layer="6032" from-port="2" to-layer="6033" to-port="1" />
-		<edge from-layer="6033" from-port="2" to-layer="6038" to-port="0" />
-		<edge from-layer="6034" from-port="0" to-layer="6037" to-port="1" />
-		<edge from-layer="6035" from-port="0" to-layer="6037" to-port="2" />
-		<edge from-layer="6036" from-port="0" to-layer="6037" to-port="3" />
-		<edge from-layer="6037" from-port="4" to-layer="6038" to-port="1" />
-		<edge from-layer="6038" from-port="2" to-layer="6039" to-port="1" />
-		<edge from-layer="6039" from-port="2" to-layer="7688" to-port="0" />
-		<edge from-layer="6039" from-port="2" to-layer="6096" to-port="1" />
-		<edge from-layer="6039" from-port="2" to-layer="6060" to-port="0" />
-		<edge from-layer="6040" from-port="0" to-layer="6041" to-port="0" />
-		<edge from-layer="6041" from-port="1" to-layer="6046" to-port="0" />
-		<edge from-layer="6042" from-port="0" to-layer="6045" to-port="1" />
-		<edge from-layer="6043" from-port="0" to-layer="6045" to-port="2" />
-		<edge from-layer="6044" from-port="0" to-layer="6045" to-port="3" />
-		<edge from-layer="6045" from-port="4" to-layer="6046" to-port="1" />
-		<edge from-layer="6046" from-port="2" to-layer="6047" to-port="0" />
-		<edge from-layer="6047" from-port="1" to-layer="6049" to-port="0" />
-		<edge from-layer="6048" from-port="0" to-layer="6049" to-port="2" />
-		<edge from-layer="6049" from-port="3" to-layer="6056" to-port="0" />
-		<edge from-layer="6050" from-port="0" to-layer="6053" to-port="1" />
-		<edge from-layer="6051" from-port="0" to-layer="6053" to-port="2" />
-		<edge from-layer="6052" from-port="0" to-layer="6053" to-port="3" />
-		<edge from-layer="6053" from-port="4" to-layer="6055" to-port="0" />
-		<edge from-layer="6054" from-port="0" to-layer="6055" to-port="1" />
-		<edge from-layer="6055" from-port="2" to-layer="6056" to-port="1" />
-		<edge from-layer="6056" from-port="2" to-layer="7687" to-port="0" />
-		<edge from-layer="6056" from-port="2" to-layer="6096" to-port="2" />
-		<edge from-layer="6057" from-port="0" to-layer="6075" to-port="0" />
-		<edge from-layer="6058" from-port="0" to-layer="6070" to-port="0" />
-		<edge from-layer="6059" from-port="0" to-layer="6070" to-port="1" />
-		<edge from-layer="6059" from-port="0" to-layer="6073" to-port="1" />
-		<edge from-layer="6060" from-port="1" to-layer="6062" to-port="0" />
-		<edge from-layer="6061" from-port="0" to-layer="6062" to-port="2" />
-		<edge from-layer="6062" from-port="3" to-layer="6072" to-port="0" />
-		<edge from-layer="6062" from-port="3" to-layer="6066" to-port="0" />
-		<edge from-layer="6063" from-port="1" to-layer="6065" to-port="0" />
-		<edge from-layer="6063" from-port="1" to-layer="6091" to-port="0" />
-		<edge from-layer="6064" from-port="0" to-layer="6065" to-port="2" />
-		<edge from-layer="6065" from-port="3" to-layer="6066" to-port="1" />
-		<edge from-layer="6066" from-port="2" to-layer="6068" to-port="0" />
-		<edge from-layer="6067" from-port="0" to-layer="6072" to-port="1" />
-		<edge from-layer="6067" from-port="0" to-layer="6068" to-port="1" />
-		<edge from-layer="6068" from-port="2" to-layer="6070" to-port="2" />
-		<edge from-layer="6069" from-port="0" to-layer="6073" to-port="3" />
-		<edge from-layer="6069" from-port="0" to-layer="6070" to-port="3" />
-		<edge from-layer="6070" from-port="4" to-layer="6075" to-port="1" />
-		<edge from-layer="6071" from-port="0" to-layer="6073" to-port="0" />
-		<edge from-layer="6072" from-port="2" to-layer="6073" to-port="2" />
-		<edge from-layer="6072" from-port="2" to-layer="6080" to-port="2" />
-		<edge from-layer="6073" from-port="4" to-layer="6075" to-port="2" />
-		<edge from-layer="6074" from-port="0" to-layer="6075" to-port="3" />
-		<edge from-layer="6075" from-port="4" to-layer="6082" to-port="0" />
-		<edge from-layer="6076" from-port="0" to-layer="6082" to-port="1" />
-		<edge from-layer="6077" from-port="0" to-layer="6080" to-port="0" />
-		<edge from-layer="6078" from-port="0" to-layer="6080" to-port="1" />
-		<edge from-layer="6079" from-port="0" to-layer="6080" to-port="3" />
-		<edge from-layer="6080" from-port="4" to-layer="6082" to-port="2" />
-		<edge from-layer="6081" from-port="0" to-layer="6082" to-port="3" />
-		<edge from-layer="6082" from-port="4" to-layer="6086" to-port="0" />
-		<edge from-layer="6083" from-port="0" to-layer="6084" to-port="0" />
-		<edge from-layer="6084" from-port="1" to-layer="6086" to-port="1" />
-		<edge from-layer="6085" from-port="0" to-layer="6086" to-port="2" />
-		<edge from-layer="6086" from-port="3" to-layer="6087" to-port="0" />
-		<edge from-layer="6087" from-port="2" to-layer="6088" to-port="0" />
-		<edge from-layer="6087" from-port="2" to-layer="6095" to-port="0" />
-		<edge from-layer="6088" from-port="1" to-layer="6094" to-port="0" />
-		<edge from-layer="6089" from-port="0" to-layer="6091" to-port="1" />
-		<edge from-layer="6090" from-port="0" to-layer="6091" to-port="2" />
-		<edge from-layer="6091" from-port="3" to-layer="6092" to-port="0" />
-		<edge from-layer="6092" from-port="4" to-layer="6093" to-port="0" />
-		<edge from-layer="6093" from-port="1" to-layer="6094" to-port="1" />
-		<edge from-layer="6094" from-port="2" to-layer="6095" to-port="1" />
-		<edge from-layer="6095" from-port="2" to-layer="6096" to-port="3" />
-		<edge from-layer="6096" from-port="4" to-layer="6098" to-port="0" />
-		<edge from-layer="6097" from-port="0" to-layer="6098" to-port="1" />
-		<edge from-layer="6098" from-port="2" to-layer="6100" to-port="0" />
-		<edge from-layer="6099" from-port="0" to-layer="6100" to-port="1" />
-		<edge from-layer="6100" from-port="2" to-layer="6109" to-port="0" />
-		<edge from-layer="6101" from-port="0" to-layer="6102" to-port="0" />
-		<edge from-layer="6102" from-port="1" to-layer="6105" to-port="0" />
-		<edge from-layer="6103" from-port="0" to-layer="6104" to-port="0" />
-		<edge from-layer="6104" from-port="1" to-layer="6105" to-port="1" />
-		<edge from-layer="6105" from-port="2" to-layer="6107" to-port="0" />
-		<edge from-layer="6106" from-port="0" to-layer="6107" to-port="1" />
-		<edge from-layer="6107" from-port="2" to-layer="6108" to-port="0" />
-		<edge from-layer="6108" from-port="1" to-layer="6109" to-port="1" />
-		<edge from-layer="6109" from-port="2" to-layer="6112" to-port="0" />
-		<edge from-layer="6110" from-port="0" to-layer="6111" to-port="0" />
-		<edge from-layer="6111" from-port="1" to-layer="6112" to-port="1" />
-		<edge from-layer="6112" from-port="2" to-layer="6113" to-port="1" />
-		<edge from-layer="6113" from-port="2" to-layer="6114" to-port="0" />
-		<edge from-layer="6114" from-port="2" to-layer="6116" to-port="0" />
-		<edge from-layer="6114" from-port="2" to-layer="6356" to-port="1" />
-		<edge from-layer="6115" from-port="0" to-layer="6116" to-port="1" />
-		<edge from-layer="6116" from-port="2" to-layer="6150" to-port="0" />
-		<edge from-layer="6116" from-port="2" to-layer="6119" to-port="0" />
-		<edge from-layer="6117" from-port="0" to-layer="6118" to-port="0" />
-		<edge from-layer="6118" from-port="1" to-layer="6119" to-port="1" />
-		<edge from-layer="6119" from-port="2" to-layer="6122" to-port="0" />
-		<edge from-layer="6120" from-port="0" to-layer="6121" to-port="0" />
-		<edge from-layer="6121" from-port="1" to-layer="6122" to-port="1" />
-		<edge from-layer="6122" from-port="2" to-layer="6131" to-port="0" />
-		<edge from-layer="6123" from-port="0" to-layer="6124" to-port="0" />
-		<edge from-layer="6124" from-port="1" to-layer="6127" to-port="0" />
-		<edge from-layer="6125" from-port="0" to-layer="6126" to-port="0" />
-		<edge from-layer="6126" from-port="1" to-layer="6127" to-port="1" />
-		<edge from-layer="6127" from-port="2" to-layer="6129" to-port="0" />
-		<edge from-layer="6128" from-port="0" to-layer="6129" to-port="1" />
-		<edge from-layer="6129" from-port="2" to-layer="6130" to-port="0" />
-		<edge from-layer="6130" from-port="1" to-layer="6131" to-port="1" />
-		<edge from-layer="6131" from-port="2" to-layer="6134" to-port="0" />
-		<edge from-layer="6132" from-port="0" to-layer="6133" to-port="0" />
-		<edge from-layer="6133" from-port="1" to-layer="6134" to-port="1" />
-		<edge from-layer="6134" from-port="2" to-layer="6135" to-port="0" />
-		<edge from-layer="6135" from-port="1" to-layer="6144" to-port="0" />
-		<edge from-layer="6136" from-port="0" to-layer="6137" to-port="0" />
-		<edge from-layer="6137" from-port="1" to-layer="6140" to-port="0" />
-		<edge from-layer="6138" from-port="0" to-layer="6139" to-port="0" />
-		<edge from-layer="6139" from-port="1" to-layer="6140" to-port="1" />
-		<edge from-layer="6140" from-port="2" to-layer="6142" to-port="0" />
-		<edge from-layer="6141" from-port="0" to-layer="6142" to-port="1" />
-		<edge from-layer="6142" from-port="2" to-layer="6143" to-port="0" />
-		<edge from-layer="6143" from-port="1" to-layer="6144" to-port="1" />
-		<edge from-layer="6144" from-port="2" to-layer="6147" to-port="0" />
-		<edge from-layer="6145" from-port="0" to-layer="6146" to-port="0" />
-		<edge from-layer="6146" from-port="1" to-layer="6147" to-port="1" />
-		<edge from-layer="6147" from-port="2" to-layer="6355" to-port="0" />
-		<edge from-layer="6148" from-port="0" to-layer="6149" to-port="0" />
-		<edge from-layer="6149" from-port="1" to-layer="6150" to-port="1" />
-		<edge from-layer="6150" from-port="2" to-layer="6153" to-port="0" />
-		<edge from-layer="6151" from-port="0" to-layer="6152" to-port="0" />
-		<edge from-layer="6152" from-port="1" to-layer="6153" to-port="1" />
-		<edge from-layer="6153" from-port="2" to-layer="6162" to-port="0" />
-		<edge from-layer="6154" from-port="0" to-layer="6155" to-port="0" />
-		<edge from-layer="6155" from-port="1" to-layer="6158" to-port="0" />
-		<edge from-layer="6156" from-port="0" to-layer="6157" to-port="0" />
-		<edge from-layer="6157" from-port="1" to-layer="6158" to-port="1" />
-		<edge from-layer="6158" from-port="2" to-layer="6160" to-port="0" />
-		<edge from-layer="6159" from-port="0" to-layer="6160" to-port="1" />
-		<edge from-layer="6160" from-port="2" to-layer="6161" to-port="0" />
-		<edge from-layer="6161" from-port="1" to-layer="6162" to-port="1" />
-		<edge from-layer="6162" from-port="2" to-layer="6165" to-port="0" />
-		<edge from-layer="6163" from-port="0" to-layer="6164" to-port="0" />
-		<edge from-layer="6164" from-port="1" to-layer="6165" to-port="1" />
-		<edge from-layer="6165" from-port="2" to-layer="6167" to-port="0" />
-		<edge from-layer="6166" from-port="0" to-layer="6167" to-port="1" />
-		<edge from-layer="6167" from-port="2" to-layer="6184" to-port="0" />
-		<edge from-layer="6167" from-port="2" to-layer="6171" to-port="0" />
-		<edge from-layer="6167" from-port="2" to-layer="6295" to-port="0" />
-		<edge from-layer="6168" from-port="0" to-layer="6171" to-port="1" />
-		<edge from-layer="6169" from-port="0" to-layer="6171" to-port="2" />
-		<edge from-layer="6170" from-port="0" to-layer="6171" to-port="3" />
-		<edge from-layer="6171" from-port="4" to-layer="6173" to-port="0" />
-		<edge from-layer="6172" from-port="0" to-layer="6173" to-port="1" />
-		<edge from-layer="6173" from-port="2" to-layer="6246" to-port="0" />
-		<edge from-layer="6173" from-port="2" to-layer="6177" to-port="0" />
-		<edge from-layer="6174" from-port="0" to-layer="6177" to-port="1" />
-		<edge from-layer="6175" from-port="0" to-layer="6177" to-port="2" />
-		<edge from-layer="6176" from-port="0" to-layer="6177" to-port="3" />
-		<edge from-layer="6177" from-port="4" to-layer="6231" to-port="0" />
-		<edge from-layer="6177" from-port="4" to-layer="6225" to-port="0" />
-		<edge from-layer="6177" from-port="4" to-layer="6214" to-port="0" />
-		<edge from-layer="6177" from-port="4" to-layer="6211" to-port="0" />
-		<edge from-layer="6178" from-port="0" to-layer="6179" to-port="0" />
-		<edge from-layer="6179" from-port="1" to-layer="6207" to-port="0" />
-		<edge from-layer="6180" from-port="0" to-layer="6207" to-port="1" />
-		<edge from-layer="6181" from-port="0" to-layer="6184" to-port="1" />
-		<edge from-layer="6182" from-port="0" to-layer="6184" to-port="2" />
-		<edge from-layer="6183" from-port="0" to-layer="6184" to-port="3" />
-		<edge from-layer="6184" from-port="4" to-layer="6186" to-port="0" />
-		<edge from-layer="6185" from-port="0" to-layer="6186" to-port="1" />
-		<edge from-layer="6186" from-port="2" to-layer="6279" to-port="0" />
-		<edge from-layer="6186" from-port="2" to-layer="6251" to-port="0" />
-		<edge from-layer="6186" from-port="2" to-layer="6187" to-port="0" />
-		<edge from-layer="6187" from-port="1" to-layer="6189" to-port="0" />
-		<edge from-layer="6188" from-port="0" to-layer="6189" to-port="2" />
-		<edge from-layer="6189" from-port="3" to-layer="6203" to-port="0" />
-		<edge from-layer="6190" from-port="0" to-layer="6191" to-port="0" />
-		<edge from-layer="6191" from-port="1" to-layer="6196" to-port="0" />
-		<edge from-layer="6192" from-port="0" to-layer="6195" to-port="1" />
-		<edge from-layer="6193" from-port="0" to-layer="6195" to-port="2" />
-		<edge from-layer="6194" from-port="0" to-layer="6195" to-port="3" />
-		<edge from-layer="6195" from-port="4" to-layer="6196" to-port="1" />
-		<edge from-layer="6196" from-port="2" to-layer="6197" to-port="0" />
-		<edge from-layer="6197" from-port="1" to-layer="6199" to-port="0" />
-		<edge from-layer="6198" from-port="0" to-layer="6199" to-port="2" />
-		<edge from-layer="6199" from-port="3" to-layer="6281" to-port="0" />
-		<edge from-layer="6199" from-port="3" to-layer="6200" to-port="0" />
-		<edge from-layer="6200" from-port="1" to-layer="6202" to-port="0" />
-		<edge from-layer="6201" from-port="0" to-layer="6202" to-port="2" />
-		<edge from-layer="6202" from-port="3" to-layer="6203" to-port="1" />
-		<edge from-layer="6203" from-port="2" to-layer="6205" to-port="0" />
-		<edge from-layer="6204" from-port="0" to-layer="6205" to-port="1" />
-		<edge from-layer="6205" from-port="2" to-layer="6207" to-port="2" />
-		<edge from-layer="6205" from-port="2" to-layer="6237" to-port="2" />
-		<edge from-layer="6206" from-port="0" to-layer="6207" to-port="3" />
-		<edge from-layer="6207" from-port="4" to-layer="6209" to-port="0" />
-		<edge from-layer="6208" from-port="0" to-layer="6209" to-port="2" />
-		<edge from-layer="6209" from-port="3" to-layer="6210" to-port="0" />
-		<edge from-layer="6210" from-port="2" to-layer="6211" to-port="1" />
-		<edge from-layer="6210" from-port="2" to-layer="6252" to-port="1" />
-		<edge from-layer="6211" from-port="2" to-layer="6242" to-port="0" />
-		<edge from-layer="6212" from-port="0" to-layer="6222" to-port="0" />
-		<edge from-layer="6213" from-port="0" to-layer="6222" to-port="1" />
-		<edge from-layer="6214" from-port="1" to-layer="6216" to-port="0" />
-		<edge from-layer="6215" from-port="0" to-layer="6216" to-port="2" />
-		<edge from-layer="6216" from-port="3" to-layer="6217" to-port="0" />
-		<edge from-layer="6217" from-port="2" to-layer="6218" to-port="0" />
-		<edge from-layer="6218" from-port="1" to-layer="6220" to-port="0" />
-		<edge from-layer="6219" from-port="0" to-layer="6220" to-port="1" />
-		<edge from-layer="6220" from-port="2" to-layer="6222" to-port="2" />
-		<edge from-layer="6221" from-port="0" to-layer="6222" to-port="3" />
-		<edge from-layer="6222" from-port="4" to-layer="6231" to-port="2" />
-		<edge from-layer="6222" from-port="4" to-layer="6225" to-port="1" />
-		<edge from-layer="6223" from-port="0" to-layer="6225" to-port="2" />
-		<edge from-layer="6224" from-port="0" to-layer="6225" to-port="3" />
-		<edge from-layer="6225" from-port="4" to-layer="6228" to-port="0" />
-		<edge from-layer="6226" from-port="0" to-layer="6227" to-port="0" />
-		<edge from-layer="6227" from-port="1" to-layer="6228" to-port="1" />
-		<edge from-layer="6228" from-port="2" to-layer="6232" to-port="0" />
-		<edge from-layer="6229" from-port="0" to-layer="6231" to-port="1" />
-		<edge from-layer="6230" from-port="0" to-layer="6231" to-port="3" />
-		<edge from-layer="6231" from-port="4" to-layer="6232" to-port="1" />
-		<edge from-layer="6232" from-port="2" to-layer="6241" to-port="0" />
-		<edge from-layer="6233" from-port="0" to-layer="6234" to-port="0" />
-		<edge from-layer="6234" from-port="1" to-layer="6237" to-port="0" />
-		<edge from-layer="6235" from-port="0" to-layer="6237" to-port="1" />
-		<edge from-layer="6236" from-port="0" to-layer="6237" to-port="3" />
-		<edge from-layer="6237" from-port="4" to-layer="6239" to-port="0" />
-		<edge from-layer="6238" from-port="0" to-layer="6239" to-port="2" />
-		<edge from-layer="6239" from-port="3" to-layer="6240" to-port="0" />
-		<edge from-layer="6240" from-port="2" to-layer="6241" to-port="1" />
-		<edge from-layer="6240" from-port="2" to-layer="6274" to-port="1" />
-		<edge from-layer="6241" from-port="2" to-layer="6242" to-port="1" />
-		<edge from-layer="6242" from-port="2" to-layer="6247" to-port="0" />
-		<edge from-layer="6243" from-port="0" to-layer="6246" to-port="1" />
-		<edge from-layer="6244" from-port="0" to-layer="6246" to-port="2" />
-		<edge from-layer="6245" from-port="0" to-layer="6246" to-port="3" />
-		<edge from-layer="6246" from-port="4" to-layer="6247" to-port="1" />
-		<edge from-layer="6247" from-port="2" to-layer="6338" to-port="0" />
-		<edge from-layer="6247" from-port="2" to-layer="6305" to-port="0" />
-		<edge from-layer="6248" from-port="0" to-layer="6251" to-port="1" />
-		<edge from-layer="6249" from-port="0" to-layer="6251" to-port="2" />
-		<edge from-layer="6250" from-port="0" to-layer="6251" to-port="3" />
-		<edge from-layer="6251" from-port="4" to-layer="6266" to-port="0" />
-		<edge from-layer="6251" from-port="4" to-layer="6252" to-port="0" />
-		<edge from-layer="6251" from-port="4" to-layer="6255" to-port="0" />
-		<edge from-layer="6251" from-port="4" to-layer="6272" to-port="0" />
-		<edge from-layer="6252" from-port="2" to-layer="6275" to-port="0" />
-		<edge from-layer="6253" from-port="0" to-layer="6263" to-port="0" />
-		<edge from-layer="6254" from-port="0" to-layer="6263" to-port="1" />
-		<edge from-layer="6255" from-port="1" to-layer="6257" to-port="0" />
-		<edge from-layer="6256" from-port="0" to-layer="6257" to-port="2" />
-		<edge from-layer="6257" from-port="3" to-layer="6258" to-port="0" />
-		<edge from-layer="6258" from-port="2" to-layer="6259" to-port="0" />
-		<edge from-layer="6259" from-port="1" to-layer="6261" to-port="0" />
-		<edge from-layer="6260" from-port="0" to-layer="6261" to-port="1" />
-		<edge from-layer="6261" from-port="2" to-layer="6263" to-port="2" />
-		<edge from-layer="6262" from-port="0" to-layer="6263" to-port="3" />
-		<edge from-layer="6263" from-port="4" to-layer="6272" to-port="2" />
-		<edge from-layer="6263" from-port="4" to-layer="6266" to-port="1" />
-		<edge from-layer="6264" from-port="0" to-layer="6266" to-port="2" />
-		<edge from-layer="6265" from-port="0" to-layer="6266" to-port="3" />
-		<edge from-layer="6266" from-port="4" to-layer="6269" to-port="0" />
-		<edge from-layer="6267" from-port="0" to-layer="6268" to-port="0" />
-		<edge from-layer="6268" from-port="1" to-layer="6269" to-port="1" />
-		<edge from-layer="6269" from-port="2" to-layer="6273" to-port="0" />
-		<edge from-layer="6270" from-port="0" to-layer="6272" to-port="1" />
-		<edge from-layer="6271" from-port="0" to-layer="6272" to-port="3" />
-		<edge from-layer="6272" from-port="4" to-layer="6273" to-port="1" />
-		<edge from-layer="6273" from-port="2" to-layer="6274" to-port="0" />
-		<edge from-layer="6274" from-port="2" to-layer="6275" to-port="1" />
-		<edge from-layer="6275" from-port="2" to-layer="6280" to-port="0" />
-		<edge from-layer="6276" from-port="0" to-layer="6279" to-port="1" />
-		<edge from-layer="6277" from-port="0" to-layer="6279" to-port="2" />
-		<edge from-layer="6278" from-port="0" to-layer="6279" to-port="3" />
-		<edge from-layer="6279" from-port="4" to-layer="6280" to-port="1" />
-		<edge from-layer="6280" from-port="2" to-layer="6281" to-port="1" />
-		<edge from-layer="6281" from-port="2" to-layer="7686" to-port="0" />
-		<edge from-layer="6281" from-port="2" to-layer="6302" to-port="0" />
-		<edge from-layer="6281" from-port="2" to-layer="6338" to-port="1" />
-		<edge from-layer="6282" from-port="0" to-layer="6283" to-port="0" />
-		<edge from-layer="6283" from-port="1" to-layer="6288" to-port="0" />
-		<edge from-layer="6284" from-port="0" to-layer="6287" to-port="1" />
-		<edge from-layer="6285" from-port="0" to-layer="6287" to-port="2" />
-		<edge from-layer="6286" from-port="0" to-layer="6287" to-port="3" />
-		<edge from-layer="6287" from-port="4" to-layer="6288" to-port="1" />
-		<edge from-layer="6288" from-port="2" to-layer="6289" to-port="0" />
-		<edge from-layer="6289" from-port="1" to-layer="6291" to-port="0" />
-		<edge from-layer="6290" from-port="0" to-layer="6291" to-port="2" />
-		<edge from-layer="6291" from-port="3" to-layer="6298" to-port="0" />
-		<edge from-layer="6292" from-port="0" to-layer="6295" to-port="1" />
-		<edge from-layer="6293" from-port="0" to-layer="6295" to-port="2" />
-		<edge from-layer="6294" from-port="0" to-layer="6295" to-port="3" />
-		<edge from-layer="6295" from-port="4" to-layer="6297" to-port="0" />
-		<edge from-layer="6296" from-port="0" to-layer="6297" to-port="1" />
-		<edge from-layer="6297" from-port="2" to-layer="6298" to-port="1" />
-		<edge from-layer="6298" from-port="2" to-layer="6338" to-port="2" />
-		<edge from-layer="6298" from-port="2" to-layer="7685" to-port="0" />
-		<edge from-layer="6299" from-port="0" to-layer="6317" to-port="0" />
-		<edge from-layer="6300" from-port="0" to-layer="6312" to-port="0" />
-		<edge from-layer="6301" from-port="0" to-layer="6312" to-port="1" />
-		<edge from-layer="6301" from-port="0" to-layer="6315" to-port="1" />
-		<edge from-layer="6302" from-port="1" to-layer="6304" to-port="0" />
-		<edge from-layer="6303" from-port="0" to-layer="6304" to-port="2" />
-		<edge from-layer="6304" from-port="3" to-layer="6314" to-port="0" />
-		<edge from-layer="6304" from-port="3" to-layer="6308" to-port="0" />
-		<edge from-layer="6305" from-port="1" to-layer="6333" to-port="0" />
-		<edge from-layer="6305" from-port="1" to-layer="6307" to-port="0" />
-		<edge from-layer="6306" from-port="0" to-layer="6307" to-port="2" />
-		<edge from-layer="6307" from-port="3" to-layer="6308" to-port="1" />
-		<edge from-layer="6308" from-port="2" to-layer="6310" to-port="0" />
-		<edge from-layer="6309" from-port="0" to-layer="6310" to-port="1" />
-		<edge from-layer="6309" from-port="0" to-layer="6314" to-port="1" />
-		<edge from-layer="6310" from-port="2" to-layer="6312" to-port="2" />
-		<edge from-layer="6311" from-port="0" to-layer="6315" to-port="3" />
-		<edge from-layer="6311" from-port="0" to-layer="6312" to-port="3" />
-		<edge from-layer="6312" from-port="4" to-layer="6317" to-port="1" />
-		<edge from-layer="6313" from-port="0" to-layer="6315" to-port="0" />
-		<edge from-layer="6314" from-port="2" to-layer="6322" to-port="2" />
-		<edge from-layer="6314" from-port="2" to-layer="6315" to-port="2" />
-		<edge from-layer="6315" from-port="4" to-layer="6317" to-port="2" />
-		<edge from-layer="6316" from-port="0" to-layer="6317" to-port="3" />
-		<edge from-layer="6317" from-port="4" to-layer="6324" to-port="0" />
-		<edge from-layer="6318" from-port="0" to-layer="6324" to-port="1" />
-		<edge from-layer="6319" from-port="0" to-layer="6322" to-port="0" />
-		<edge from-layer="6320" from-port="0" to-layer="6322" to-port="1" />
-		<edge from-layer="6321" from-port="0" to-layer="6322" to-port="3" />
-		<edge from-layer="6322" from-port="4" to-layer="6324" to-port="2" />
-		<edge from-layer="6323" from-port="0" to-layer="6324" to-port="3" />
-		<edge from-layer="6324" from-port="4" to-layer="6328" to-port="0" />
-		<edge from-layer="6325" from-port="0" to-layer="6326" to-port="0" />
-		<edge from-layer="6326" from-port="1" to-layer="6328" to-port="1" />
-		<edge from-layer="6327" from-port="0" to-layer="6328" to-port="2" />
-		<edge from-layer="6328" from-port="3" to-layer="6329" to-port="0" />
-		<edge from-layer="6329" from-port="2" to-layer="6330" to-port="0" />
-		<edge from-layer="6329" from-port="2" to-layer="6337" to-port="0" />
-		<edge from-layer="6330" from-port="1" to-layer="6336" to-port="0" />
-		<edge from-layer="6331" from-port="0" to-layer="6333" to-port="1" />
-		<edge from-layer="6332" from-port="0" to-layer="6333" to-port="2" />
-		<edge from-layer="6333" from-port="3" to-layer="6334" to-port="0" />
-		<edge from-layer="6334" from-port="4" to-layer="6335" to-port="0" />
-		<edge from-layer="6335" from-port="1" to-layer="6336" to-port="1" />
-		<edge from-layer="6336" from-port="2" to-layer="6337" to-port="1" />
-		<edge from-layer="6337" from-port="2" to-layer="6338" to-port="3" />
-		<edge from-layer="6338" from-port="4" to-layer="6340" to-port="0" />
-		<edge from-layer="6339" from-port="0" to-layer="6340" to-port="1" />
-		<edge from-layer="6340" from-port="2" to-layer="6342" to-port="0" />
-		<edge from-layer="6341" from-port="0" to-layer="6342" to-port="1" />
-		<edge from-layer="6342" from-port="2" to-layer="6351" to-port="0" />
-		<edge from-layer="6343" from-port="0" to-layer="6344" to-port="0" />
-		<edge from-layer="6344" from-port="1" to-layer="6347" to-port="0" />
-		<edge from-layer="6345" from-port="0" to-layer="6346" to-port="0" />
-		<edge from-layer="6346" from-port="1" to-layer="6347" to-port="1" />
-		<edge from-layer="6347" from-port="2" to-layer="6349" to-port="0" />
-		<edge from-layer="6348" from-port="0" to-layer="6349" to-port="1" />
-		<edge from-layer="6349" from-port="2" to-layer="6350" to-port="0" />
-		<edge from-layer="6350" from-port="1" to-layer="6351" to-port="1" />
-		<edge from-layer="6351" from-port="2" to-layer="6354" to-port="0" />
-		<edge from-layer="6352" from-port="0" to-layer="6353" to-port="0" />
-		<edge from-layer="6353" from-port="1" to-layer="6354" to-port="1" />
-		<edge from-layer="6354" from-port="2" to-layer="6355" to-port="1" />
-		<edge from-layer="6355" from-port="2" to-layer="6356" to-port="0" />
-		<edge from-layer="6356" from-port="2" to-layer="6598" to-port="1" />
-		<edge from-layer="6356" from-port="2" to-layer="6358" to-port="0" />
-		<edge from-layer="6357" from-port="0" to-layer="6358" to-port="1" />
-		<edge from-layer="6358" from-port="2" to-layer="6361" to-port="0" />
-		<edge from-layer="6358" from-port="2" to-layer="6392" to-port="0" />
-		<edge from-layer="6359" from-port="0" to-layer="6360" to-port="0" />
-		<edge from-layer="6360" from-port="1" to-layer="6361" to-port="1" />
-		<edge from-layer="6361" from-port="2" to-layer="6364" to-port="0" />
-		<edge from-layer="6362" from-port="0" to-layer="6363" to-port="0" />
-		<edge from-layer="6363" from-port="1" to-layer="6364" to-port="1" />
-		<edge from-layer="6364" from-port="2" to-layer="6373" to-port="0" />
-		<edge from-layer="6365" from-port="0" to-layer="6366" to-port="0" />
-		<edge from-layer="6366" from-port="1" to-layer="6369" to-port="0" />
-		<edge from-layer="6367" from-port="0" to-layer="6368" to-port="0" />
-		<edge from-layer="6368" from-port="1" to-layer="6369" to-port="1" />
-		<edge from-layer="6369" from-port="2" to-layer="6371" to-port="0" />
-		<edge from-layer="6370" from-port="0" to-layer="6371" to-port="1" />
-		<edge from-layer="6371" from-port="2" to-layer="6372" to-port="0" />
-		<edge from-layer="6372" from-port="1" to-layer="6373" to-port="1" />
-		<edge from-layer="6373" from-port="2" to-layer="6376" to-port="0" />
-		<edge from-layer="6374" from-port="0" to-layer="6375" to-port="0" />
-		<edge from-layer="6375" from-port="1" to-layer="6376" to-port="1" />
-		<edge from-layer="6376" from-port="2" to-layer="6377" to-port="0" />
-		<edge from-layer="6377" from-port="1" to-layer="6386" to-port="0" />
-		<edge from-layer="6378" from-port="0" to-layer="6379" to-port="0" />
-		<edge from-layer="6379" from-port="1" to-layer="6382" to-port="0" />
-		<edge from-layer="6380" from-port="0" to-layer="6381" to-port="0" />
-		<edge from-layer="6381" from-port="1" to-layer="6382" to-port="1" />
-		<edge from-layer="6382" from-port="2" to-layer="6384" to-port="0" />
-		<edge from-layer="6383" from-port="0" to-layer="6384" to-port="1" />
-		<edge from-layer="6384" from-port="2" to-layer="6385" to-port="0" />
-		<edge from-layer="6385" from-port="1" to-layer="6386" to-port="1" />
-		<edge from-layer="6386" from-port="2" to-layer="6389" to-port="0" />
-		<edge from-layer="6387" from-port="0" to-layer="6388" to-port="0" />
-		<edge from-layer="6388" from-port="1" to-layer="6389" to-port="1" />
-		<edge from-layer="6389" from-port="2" to-layer="6597" to-port="0" />
-		<edge from-layer="6390" from-port="0" to-layer="6391" to-port="0" />
-		<edge from-layer="6391" from-port="1" to-layer="6392" to-port="1" />
-		<edge from-layer="6392" from-port="2" to-layer="6395" to-port="0" />
-		<edge from-layer="6393" from-port="0" to-layer="6394" to-port="0" />
-		<edge from-layer="6394" from-port="1" to-layer="6395" to-port="1" />
-		<edge from-layer="6395" from-port="2" to-layer="6404" to-port="0" />
-		<edge from-layer="6396" from-port="0" to-layer="6397" to-port="0" />
-		<edge from-layer="6397" from-port="1" to-layer="6400" to-port="0" />
-		<edge from-layer="6398" from-port="0" to-layer="6399" to-port="0" />
-		<edge from-layer="6399" from-port="1" to-layer="6400" to-port="1" />
-		<edge from-layer="6400" from-port="2" to-layer="6402" to-port="0" />
-		<edge from-layer="6401" from-port="0" to-layer="6402" to-port="1" />
-		<edge from-layer="6402" from-port="2" to-layer="6403" to-port="0" />
-		<edge from-layer="6403" from-port="1" to-layer="6404" to-port="1" />
-		<edge from-layer="6404" from-port="2" to-layer="6407" to-port="0" />
-		<edge from-layer="6405" from-port="0" to-layer="6406" to-port="0" />
-		<edge from-layer="6406" from-port="1" to-layer="6407" to-port="1" />
-		<edge from-layer="6407" from-port="2" to-layer="6409" to-port="0" />
-		<edge from-layer="6408" from-port="0" to-layer="6409" to-port="1" />
-		<edge from-layer="6409" from-port="2" to-layer="6537" to-port="0" />
-		<edge from-layer="6409" from-port="2" to-layer="6413" to-port="0" />
-		<edge from-layer="6409" from-port="2" to-layer="6426" to-port="0" />
-		<edge from-layer="6410" from-port="0" to-layer="6413" to-port="1" />
-		<edge from-layer="6411" from-port="0" to-layer="6413" to-port="2" />
-		<edge from-layer="6412" from-port="0" to-layer="6413" to-port="3" />
-		<edge from-layer="6413" from-port="4" to-layer="6415" to-port="0" />
-		<edge from-layer="6414" from-port="0" to-layer="6415" to-port="1" />
-		<edge from-layer="6415" from-port="2" to-layer="6488" to-port="0" />
-		<edge from-layer="6415" from-port="2" to-layer="6419" to-port="0" />
-		<edge from-layer="6416" from-port="0" to-layer="6419" to-port="1" />
-		<edge from-layer="6417" from-port="0" to-layer="6419" to-port="2" />
-		<edge from-layer="6418" from-port="0" to-layer="6419" to-port="3" />
-		<edge from-layer="6419" from-port="4" to-layer="6456" to-port="0" />
-		<edge from-layer="6419" from-port="4" to-layer="6453" to-port="0" />
-		<edge from-layer="6419" from-port="4" to-layer="6467" to-port="0" />
-		<edge from-layer="6419" from-port="4" to-layer="6473" to-port="0" />
-		<edge from-layer="6420" from-port="0" to-layer="6421" to-port="0" />
-		<edge from-layer="6421" from-port="1" to-layer="6449" to-port="0" />
-		<edge from-layer="6422" from-port="0" to-layer="6449" to-port="1" />
-		<edge from-layer="6423" from-port="0" to-layer="6426" to-port="1" />
-		<edge from-layer="6424" from-port="0" to-layer="6426" to-port="2" />
-		<edge from-layer="6425" from-port="0" to-layer="6426" to-port="3" />
-		<edge from-layer="6426" from-port="4" to-layer="6428" to-port="0" />
-		<edge from-layer="6427" from-port="0" to-layer="6428" to-port="1" />
-		<edge from-layer="6428" from-port="2" to-layer="6429" to-port="0" />
-		<edge from-layer="6428" from-port="2" to-layer="6521" to-port="0" />
-		<edge from-layer="6428" from-port="2" to-layer="6493" to-port="0" />
-		<edge from-layer="6429" from-port="1" to-layer="6431" to-port="0" />
-		<edge from-layer="6430" from-port="0" to-layer="6431" to-port="2" />
-		<edge from-layer="6431" from-port="3" to-layer="6445" to-port="0" />
-		<edge from-layer="6432" from-port="0" to-layer="6433" to-port="0" />
-		<edge from-layer="6433" from-port="1" to-layer="6438" to-port="0" />
-		<edge from-layer="6434" from-port="0" to-layer="6437" to-port="1" />
-		<edge from-layer="6435" from-port="0" to-layer="6437" to-port="2" />
-		<edge from-layer="6436" from-port="0" to-layer="6437" to-port="3" />
-		<edge from-layer="6437" from-port="4" to-layer="6438" to-port="1" />
-		<edge from-layer="6438" from-port="2" to-layer="6439" to-port="0" />
-		<edge from-layer="6439" from-port="1" to-layer="6441" to-port="0" />
-		<edge from-layer="6440" from-port="0" to-layer="6441" to-port="2" />
-		<edge from-layer="6441" from-port="3" to-layer="6442" to-port="0" />
-		<edge from-layer="6441" from-port="3" to-layer="6523" to-port="0" />
-		<edge from-layer="6442" from-port="1" to-layer="6444" to-port="0" />
-		<edge from-layer="6443" from-port="0" to-layer="6444" to-port="2" />
-		<edge from-layer="6444" from-port="3" to-layer="6445" to-port="1" />
-		<edge from-layer="6445" from-port="2" to-layer="6447" to-port="0" />
-		<edge from-layer="6446" from-port="0" to-layer="6447" to-port="1" />
-		<edge from-layer="6447" from-port="2" to-layer="6479" to-port="2" />
-		<edge from-layer="6447" from-port="2" to-layer="6449" to-port="2" />
-		<edge from-layer="6448" from-port="0" to-layer="6449" to-port="3" />
-		<edge from-layer="6449" from-port="4" to-layer="6451" to-port="0" />
-		<edge from-layer="6450" from-port="0" to-layer="6451" to-port="2" />
-		<edge from-layer="6451" from-port="3" to-layer="6452" to-port="0" />
-		<edge from-layer="6452" from-port="2" to-layer="6494" to-port="1" />
-		<edge from-layer="6452" from-port="2" to-layer="6453" to-port="1" />
-		<edge from-layer="6453" from-port="2" to-layer="6484" to-port="0" />
-		<edge from-layer="6454" from-port="0" to-layer="6464" to-port="0" />
-		<edge from-layer="6455" from-port="0" to-layer="6464" to-port="1" />
-		<edge from-layer="6456" from-port="1" to-layer="6458" to-port="0" />
-		<edge from-layer="6457" from-port="0" to-layer="6458" to-port="2" />
-		<edge from-layer="6458" from-port="3" to-layer="6459" to-port="0" />
-		<edge from-layer="6459" from-port="2" to-layer="6460" to-port="0" />
-		<edge from-layer="6460" from-port="1" to-layer="6462" to-port="0" />
-		<edge from-layer="6461" from-port="0" to-layer="6462" to-port="1" />
-		<edge from-layer="6462" from-port="2" to-layer="6464" to-port="2" />
-		<edge from-layer="6463" from-port="0" to-layer="6464" to-port="3" />
-		<edge from-layer="6464" from-port="4" to-layer="6473" to-port="2" />
-		<edge from-layer="6464" from-port="4" to-layer="6467" to-port="1" />
-		<edge from-layer="6465" from-port="0" to-layer="6467" to-port="2" />
-		<edge from-layer="6466" from-port="0" to-layer="6467" to-port="3" />
-		<edge from-layer="6467" from-port="4" to-layer="6470" to-port="0" />
-		<edge from-layer="6468" from-port="0" to-layer="6469" to-port="0" />
-		<edge from-layer="6469" from-port="1" to-layer="6470" to-port="1" />
-		<edge from-layer="6470" from-port="2" to-layer="6474" to-port="0" />
-		<edge from-layer="6471" from-port="0" to-layer="6473" to-port="1" />
-		<edge from-layer="6472" from-port="0" to-layer="6473" to-port="3" />
-		<edge from-layer="6473" from-port="4" to-layer="6474" to-port="1" />
-		<edge from-layer="6474" from-port="2" to-layer="6483" to-port="0" />
-		<edge from-layer="6475" from-port="0" to-layer="6476" to-port="0" />
-		<edge from-layer="6476" from-port="1" to-layer="6479" to-port="0" />
-		<edge from-layer="6477" from-port="0" to-layer="6479" to-port="1" />
-		<edge from-layer="6478" from-port="0" to-layer="6479" to-port="3" />
-		<edge from-layer="6479" from-port="4" to-layer="6481" to-port="0" />
-		<edge from-layer="6480" from-port="0" to-layer="6481" to-port="2" />
-		<edge from-layer="6481" from-port="3" to-layer="6482" to-port="0" />
-		<edge from-layer="6482" from-port="2" to-layer="6483" to-port="1" />
-		<edge from-layer="6482" from-port="2" to-layer="6516" to-port="1" />
-		<edge from-layer="6483" from-port="2" to-layer="6484" to-port="1" />
-		<edge from-layer="6484" from-port="2" to-layer="6489" to-port="0" />
-		<edge from-layer="6485" from-port="0" to-layer="6488" to-port="1" />
-		<edge from-layer="6486" from-port="0" to-layer="6488" to-port="2" />
-		<edge from-layer="6487" from-port="0" to-layer="6488" to-port="3" />
-		<edge from-layer="6488" from-port="4" to-layer="6489" to-port="1" />
-		<edge from-layer="6489" from-port="2" to-layer="6547" to-port="0" />
-		<edge from-layer="6489" from-port="2" to-layer="6580" to-port="0" />
-		<edge from-layer="6490" from-port="0" to-layer="6493" to-port="1" />
-		<edge from-layer="6491" from-port="0" to-layer="6493" to-port="2" />
-		<edge from-layer="6492" from-port="0" to-layer="6493" to-port="3" />
-		<edge from-layer="6493" from-port="4" to-layer="6494" to-port="0" />
-		<edge from-layer="6493" from-port="4" to-layer="6497" to-port="0" />
-		<edge from-layer="6493" from-port="4" to-layer="6514" to-port="0" />
-		<edge from-layer="6493" from-port="4" to-layer="6508" to-port="0" />
-		<edge from-layer="6494" from-port="2" to-layer="6517" to-port="0" />
-		<edge from-layer="6495" from-port="0" to-layer="6505" to-port="0" />
-		<edge from-layer="6496" from-port="0" to-layer="6505" to-port="1" />
-		<edge from-layer="6497" from-port="1" to-layer="6499" to-port="0" />
-		<edge from-layer="6498" from-port="0" to-layer="6499" to-port="2" />
-		<edge from-layer="6499" from-port="3" to-layer="6500" to-port="0" />
-		<edge from-layer="6500" from-port="2" to-layer="6501" to-port="0" />
-		<edge from-layer="6501" from-port="1" to-layer="6503" to-port="0" />
-		<edge from-layer="6502" from-port="0" to-layer="6503" to-port="1" />
-		<edge from-layer="6503" from-port="2" to-layer="6505" to-port="2" />
-		<edge from-layer="6504" from-port="0" to-layer="6505" to-port="3" />
-		<edge from-layer="6505" from-port="4" to-layer="6508" to-port="1" />
-		<edge from-layer="6505" from-port="4" to-layer="6514" to-port="2" />
-		<edge from-layer="6506" from-port="0" to-layer="6508" to-port="2" />
-		<edge from-layer="6507" from-port="0" to-layer="6508" to-port="3" />
-		<edge from-layer="6508" from-port="4" to-layer="6511" to-port="0" />
-		<edge from-layer="6509" from-port="0" to-layer="6510" to-port="0" />
-		<edge from-layer="6510" from-port="1" to-layer="6511" to-port="1" />
-		<edge from-layer="6511" from-port="2" to-layer="6515" to-port="0" />
-		<edge from-layer="6512" from-port="0" to-layer="6514" to-port="1" />
-		<edge from-layer="6513" from-port="0" to-layer="6514" to-port="3" />
-		<edge from-layer="6514" from-port="4" to-layer="6515" to-port="1" />
-		<edge from-layer="6515" from-port="2" to-layer="6516" to-port="0" />
-		<edge from-layer="6516" from-port="2" to-layer="6517" to-port="1" />
-		<edge from-layer="6517" from-port="2" to-layer="6522" to-port="0" />
-		<edge from-layer="6518" from-port="0" to-layer="6521" to-port="1" />
-		<edge from-layer="6519" from-port="0" to-layer="6521" to-port="2" />
-		<edge from-layer="6520" from-port="0" to-layer="6521" to-port="3" />
-		<edge from-layer="6521" from-port="4" to-layer="6522" to-port="1" />
-		<edge from-layer="6522" from-port="2" to-layer="6523" to-port="1" />
-		<edge from-layer="6523" from-port="2" to-layer="6544" to-port="0" />
-		<edge from-layer="6523" from-port="2" to-layer="6580" to-port="1" />
-		<edge from-layer="6523" from-port="2" to-layer="7684" to-port="0" />
-		<edge from-layer="6524" from-port="0" to-layer="6525" to-port="0" />
-		<edge from-layer="6525" from-port="1" to-layer="6530" to-port="0" />
-		<edge from-layer="6526" from-port="0" to-layer="6529" to-port="1" />
-		<edge from-layer="6527" from-port="0" to-layer="6529" to-port="2" />
-		<edge from-layer="6528" from-port="0" to-layer="6529" to-port="3" />
-		<edge from-layer="6529" from-port="4" to-layer="6530" to-port="1" />
-		<edge from-layer="6530" from-port="2" to-layer="6531" to-port="0" />
-		<edge from-layer="6531" from-port="1" to-layer="6533" to-port="0" />
-		<edge from-layer="6532" from-port="0" to-layer="6533" to-port="2" />
-		<edge from-layer="6533" from-port="3" to-layer="6540" to-port="0" />
-		<edge from-layer="6534" from-port="0" to-layer="6537" to-port="1" />
-		<edge from-layer="6535" from-port="0" to-layer="6537" to-port="2" />
-		<edge from-layer="6536" from-port="0" to-layer="6537" to-port="3" />
-		<edge from-layer="6537" from-port="4" to-layer="6539" to-port="0" />
-		<edge from-layer="6538" from-port="0" to-layer="6539" to-port="1" />
-		<edge from-layer="6539" from-port="2" to-layer="6540" to-port="1" />
-		<edge from-layer="6540" from-port="2" to-layer="7683" to-port="0" />
-		<edge from-layer="6540" from-port="2" to-layer="6580" to-port="2" />
-		<edge from-layer="6541" from-port="0" to-layer="6559" to-port="0" />
-		<edge from-layer="6542" from-port="0" to-layer="6554" to-port="0" />
-		<edge from-layer="6543" from-port="0" to-layer="6554" to-port="1" />
-		<edge from-layer="6543" from-port="0" to-layer="6557" to-port="1" />
-		<edge from-layer="6544" from-port="1" to-layer="6546" to-port="0" />
-		<edge from-layer="6545" from-port="0" to-layer="6546" to-port="2" />
-		<edge from-layer="6546" from-port="3" to-layer="6550" to-port="0" />
-		<edge from-layer="6546" from-port="3" to-layer="6556" to-port="0" />
-		<edge from-layer="6547" from-port="1" to-layer="6575" to-port="0" />
-		<edge from-layer="6547" from-port="1" to-layer="6549" to-port="0" />
-		<edge from-layer="6548" from-port="0" to-layer="6549" to-port="2" />
-		<edge from-layer="6549" from-port="3" to-layer="6550" to-port="1" />
-		<edge from-layer="6550" from-port="2" to-layer="6552" to-port="0" />
-		<edge from-layer="6551" from-port="0" to-layer="6552" to-port="1" />
-		<edge from-layer="6551" from-port="0" to-layer="6556" to-port="1" />
-		<edge from-layer="6552" from-port="2" to-layer="6554" to-port="2" />
-		<edge from-layer="6553" from-port="0" to-layer="6554" to-port="3" />
-		<edge from-layer="6553" from-port="0" to-layer="6557" to-port="3" />
-		<edge from-layer="6554" from-port="4" to-layer="6559" to-port="1" />
-		<edge from-layer="6555" from-port="0" to-layer="6557" to-port="0" />
-		<edge from-layer="6556" from-port="2" to-layer="6557" to-port="2" />
-		<edge from-layer="6556" from-port="2" to-layer="6564" to-port="2" />
-		<edge from-layer="6557" from-port="4" to-layer="6559" to-port="2" />
-		<edge from-layer="6558" from-port="0" to-layer="6559" to-port="3" />
-		<edge from-layer="6559" from-port="4" to-layer="6566" to-port="0" />
-		<edge from-layer="6560" from-port="0" to-layer="6566" to-port="1" />
-		<edge from-layer="6561" from-port="0" to-layer="6564" to-port="0" />
-		<edge from-layer="6562" from-port="0" to-layer="6564" to-port="1" />
-		<edge from-layer="6563" from-port="0" to-layer="6564" to-port="3" />
-		<edge from-layer="6564" from-port="4" to-layer="6566" to-port="2" />
-		<edge from-layer="6565" from-port="0" to-layer="6566" to-port="3" />
-		<edge from-layer="6566" from-port="4" to-layer="6570" to-port="0" />
-		<edge from-layer="6567" from-port="0" to-layer="6568" to-port="0" />
-		<edge from-layer="6568" from-port="1" to-layer="6570" to-port="1" />
-		<edge from-layer="6569" from-port="0" to-layer="6570" to-port="2" />
-		<edge from-layer="6570" from-port="3" to-layer="6571" to-port="0" />
-		<edge from-layer="6571" from-port="2" to-layer="6579" to-port="0" />
-		<edge from-layer="6571" from-port="2" to-layer="6572" to-port="0" />
-		<edge from-layer="6572" from-port="1" to-layer="6578" to-port="0" />
-		<edge from-layer="6573" from-port="0" to-layer="6575" to-port="1" />
-		<edge from-layer="6574" from-port="0" to-layer="6575" to-port="2" />
-		<edge from-layer="6575" from-port="3" to-layer="6576" to-port="0" />
-		<edge from-layer="6576" from-port="4" to-layer="6577" to-port="0" />
-		<edge from-layer="6577" from-port="1" to-layer="6578" to-port="1" />
-		<edge from-layer="6578" from-port="2" to-layer="6579" to-port="1" />
-		<edge from-layer="6579" from-port="2" to-layer="6580" to-port="3" />
-		<edge from-layer="6580" from-port="4" to-layer="6582" to-port="0" />
-		<edge from-layer="6581" from-port="0" to-layer="6582" to-port="1" />
-		<edge from-layer="6582" from-port="2" to-layer="6584" to-port="0" />
-		<edge from-layer="6583" from-port="0" to-layer="6584" to-port="1" />
-		<edge from-layer="6584" from-port="2" to-layer="6593" to-port="0" />
-		<edge from-layer="6585" from-port="0" to-layer="6586" to-port="0" />
-		<edge from-layer="6586" from-port="1" to-layer="6589" to-port="0" />
-		<edge from-layer="6587" from-port="0" to-layer="6588" to-port="0" />
-		<edge from-layer="6588" from-port="1" to-layer="6589" to-port="1" />
-		<edge from-layer="6589" from-port="2" to-layer="6591" to-port="0" />
-		<edge from-layer="6590" from-port="0" to-layer="6591" to-port="1" />
-		<edge from-layer="6591" from-port="2" to-layer="6592" to-port="0" />
-		<edge from-layer="6592" from-port="1" to-layer="6593" to-port="1" />
-		<edge from-layer="6593" from-port="2" to-layer="6596" to-port="0" />
-		<edge from-layer="6594" from-port="0" to-layer="6595" to-port="0" />
-		<edge from-layer="6595" from-port="1" to-layer="6596" to-port="1" />
-		<edge from-layer="6596" from-port="2" to-layer="6597" to-port="1" />
-		<edge from-layer="6597" from-port="2" to-layer="6598" to-port="0" />
-		<edge from-layer="6598" from-port="2" to-layer="6840" to-port="1" />
-		<edge from-layer="6598" from-port="2" to-layer="6600" to-port="0" />
-		<edge from-layer="6599" from-port="0" to-layer="6600" to-port="1" />
-		<edge from-layer="6600" from-port="2" to-layer="6634" to-port="0" />
-		<edge from-layer="6600" from-port="2" to-layer="6603" to-port="0" />
-		<edge from-layer="6601" from-port="0" to-layer="6602" to-port="0" />
-		<edge from-layer="6602" from-port="1" to-layer="6603" to-port="1" />
-		<edge from-layer="6603" from-port="2" to-layer="6606" to-port="0" />
-		<edge from-layer="6604" from-port="0" to-layer="6605" to-port="0" />
-		<edge from-layer="6605" from-port="1" to-layer="6606" to-port="1" />
-		<edge from-layer="6606" from-port="2" to-layer="6615" to-port="0" />
-		<edge from-layer="6607" from-port="0" to-layer="6608" to-port="0" />
-		<edge from-layer="6608" from-port="1" to-layer="6611" to-port="0" />
-		<edge from-layer="6609" from-port="0" to-layer="6610" to-port="0" />
-		<edge from-layer="6610" from-port="1" to-layer="6611" to-port="1" />
-		<edge from-layer="6611" from-port="2" to-layer="6613" to-port="0" />
-		<edge from-layer="6612" from-port="0" to-layer="6613" to-port="1" />
-		<edge from-layer="6613" from-port="2" to-layer="6614" to-port="0" />
-		<edge from-layer="6614" from-port="1" to-layer="6615" to-port="1" />
-		<edge from-layer="6615" from-port="2" to-layer="6618" to-port="0" />
-		<edge from-layer="6616" from-port="0" to-layer="6617" to-port="0" />
-		<edge from-layer="6617" from-port="1" to-layer="6618" to-port="1" />
-		<edge from-layer="6618" from-port="2" to-layer="6619" to-port="0" />
-		<edge from-layer="6619" from-port="1" to-layer="6628" to-port="0" />
-		<edge from-layer="6620" from-port="0" to-layer="6621" to-port="0" />
-		<edge from-layer="6621" from-port="1" to-layer="6624" to-port="0" />
-		<edge from-layer="6622" from-port="0" to-layer="6623" to-port="0" />
-		<edge from-layer="6623" from-port="1" to-layer="6624" to-port="1" />
-		<edge from-layer="6624" from-port="2" to-layer="6626" to-port="0" />
-		<edge from-layer="6625" from-port="0" to-layer="6626" to-port="1" />
-		<edge from-layer="6626" from-port="2" to-layer="6627" to-port="0" />
-		<edge from-layer="6627" from-port="1" to-layer="6628" to-port="1" />
-		<edge from-layer="6628" from-port="2" to-layer="6631" to-port="0" />
-		<edge from-layer="6629" from-port="0" to-layer="6630" to-port="0" />
-		<edge from-layer="6630" from-port="1" to-layer="6631" to-port="1" />
-		<edge from-layer="6631" from-port="2" to-layer="6839" to-port="0" />
-		<edge from-layer="6632" from-port="0" to-layer="6633" to-port="0" />
-		<edge from-layer="6633" from-port="1" to-layer="6634" to-port="1" />
-		<edge from-layer="6634" from-port="2" to-layer="6637" to-port="0" />
-		<edge from-layer="6635" from-port="0" to-layer="6636" to-port="0" />
-		<edge from-layer="6636" from-port="1" to-layer="6637" to-port="1" />
-		<edge from-layer="6637" from-port="2" to-layer="6646" to-port="0" />
-		<edge from-layer="6638" from-port="0" to-layer="6639" to-port="0" />
-		<edge from-layer="6639" from-port="1" to-layer="6642" to-port="0" />
-		<edge from-layer="6640" from-port="0" to-layer="6641" to-port="0" />
-		<edge from-layer="6641" from-port="1" to-layer="6642" to-port="1" />
-		<edge from-layer="6642" from-port="2" to-layer="6644" to-port="0" />
-		<edge from-layer="6643" from-port="0" to-layer="6644" to-port="1" />
-		<edge from-layer="6644" from-port="2" to-layer="6645" to-port="0" />
-		<edge from-layer="6645" from-port="1" to-layer="6646" to-port="1" />
-		<edge from-layer="6646" from-port="2" to-layer="6649" to-port="0" />
-		<edge from-layer="6647" from-port="0" to-layer="6648" to-port="0" />
-		<edge from-layer="6648" from-port="1" to-layer="6649" to-port="1" />
-		<edge from-layer="6649" from-port="2" to-layer="6651" to-port="0" />
-		<edge from-layer="6650" from-port="0" to-layer="6651" to-port="1" />
-		<edge from-layer="6651" from-port="2" to-layer="6655" to-port="0" />
-		<edge from-layer="6651" from-port="2" to-layer="6779" to-port="0" />
-		<edge from-layer="6651" from-port="2" to-layer="6668" to-port="0" />
-		<edge from-layer="6652" from-port="0" to-layer="6655" to-port="1" />
-		<edge from-layer="6653" from-port="0" to-layer="6655" to-port="2" />
-		<edge from-layer="6654" from-port="0" to-layer="6655" to-port="3" />
-		<edge from-layer="6655" from-port="4" to-layer="6657" to-port="0" />
-		<edge from-layer="6656" from-port="0" to-layer="6657" to-port="1" />
-		<edge from-layer="6657" from-port="2" to-layer="6661" to-port="0" />
-		<edge from-layer="6657" from-port="2" to-layer="6730" to-port="0" />
-		<edge from-layer="6658" from-port="0" to-layer="6661" to-port="1" />
-		<edge from-layer="6659" from-port="0" to-layer="6661" to-port="2" />
-		<edge from-layer="6660" from-port="0" to-layer="6661" to-port="3" />
-		<edge from-layer="6661" from-port="4" to-layer="6715" to-port="0" />
-		<edge from-layer="6661" from-port="4" to-layer="6695" to-port="0" />
-		<edge from-layer="6661" from-port="4" to-layer="6709" to-port="0" />
-		<edge from-layer="6661" from-port="4" to-layer="6698" to-port="0" />
-		<edge from-layer="6662" from-port="0" to-layer="6663" to-port="0" />
-		<edge from-layer="6663" from-port="1" to-layer="6691" to-port="0" />
-		<edge from-layer="6664" from-port="0" to-layer="6691" to-port="1" />
-		<edge from-layer="6665" from-port="0" to-layer="6668" to-port="1" />
-		<edge from-layer="6666" from-port="0" to-layer="6668" to-port="2" />
-		<edge from-layer="6667" from-port="0" to-layer="6668" to-port="3" />
-		<edge from-layer="6668" from-port="4" to-layer="6670" to-port="0" />
-		<edge from-layer="6669" from-port="0" to-layer="6670" to-port="1" />
-		<edge from-layer="6670" from-port="2" to-layer="6735" to-port="0" />
-		<edge from-layer="6670" from-port="2" to-layer="6763" to-port="0" />
-		<edge from-layer="6670" from-port="2" to-layer="6671" to-port="0" />
-		<edge from-layer="6671" from-port="1" to-layer="6673" to-port="0" />
-		<edge from-layer="6672" from-port="0" to-layer="6673" to-port="2" />
-		<edge from-layer="6673" from-port="3" to-layer="6687" to-port="0" />
-		<edge from-layer="6674" from-port="0" to-layer="6675" to-port="0" />
-		<edge from-layer="6675" from-port="1" to-layer="6680" to-port="0" />
-		<edge from-layer="6676" from-port="0" to-layer="6679" to-port="1" />
-		<edge from-layer="6677" from-port="0" to-layer="6679" to-port="2" />
-		<edge from-layer="6678" from-port="0" to-layer="6679" to-port="3" />
-		<edge from-layer="6679" from-port="4" to-layer="6680" to-port="1" />
-		<edge from-layer="6680" from-port="2" to-layer="6681" to-port="0" />
-		<edge from-layer="6681" from-port="1" to-layer="6683" to-port="0" />
-		<edge from-layer="6682" from-port="0" to-layer="6683" to-port="2" />
-		<edge from-layer="6683" from-port="3" to-layer="6765" to-port="0" />
-		<edge from-layer="6683" from-port="3" to-layer="6684" to-port="0" />
-		<edge from-layer="6684" from-port="1" to-layer="6686" to-port="0" />
-		<edge from-layer="6685" from-port="0" to-layer="6686" to-port="2" />
-		<edge from-layer="6686" from-port="3" to-layer="6687" to-port="1" />
-		<edge from-layer="6687" from-port="2" to-layer="6689" to-port="0" />
-		<edge from-layer="6688" from-port="0" to-layer="6689" to-port="1" />
-		<edge from-layer="6689" from-port="2" to-layer="6691" to-port="2" />
-		<edge from-layer="6689" from-port="2" to-layer="6721" to-port="2" />
-		<edge from-layer="6690" from-port="0" to-layer="6691" to-port="3" />
-		<edge from-layer="6691" from-port="4" to-layer="6693" to-port="0" />
-		<edge from-layer="6692" from-port="0" to-layer="6693" to-port="2" />
-		<edge from-layer="6693" from-port="3" to-layer="6694" to-port="0" />
-		<edge from-layer="6694" from-port="2" to-layer="6736" to-port="1" />
-		<edge from-layer="6694" from-port="2" to-layer="6695" to-port="1" />
-		<edge from-layer="6695" from-port="2" to-layer="6726" to-port="0" />
-		<edge from-layer="6696" from-port="0" to-layer="6706" to-port="0" />
-		<edge from-layer="6697" from-port="0" to-layer="6706" to-port="1" />
-		<edge from-layer="6698" from-port="1" to-layer="6700" to-port="0" />
-		<edge from-layer="6699" from-port="0" to-layer="6700" to-port="2" />
-		<edge from-layer="6700" from-port="3" to-layer="6701" to-port="0" />
-		<edge from-layer="6701" from-port="2" to-layer="6702" to-port="0" />
-		<edge from-layer="6702" from-port="1" to-layer="6704" to-port="0" />
-		<edge from-layer="6703" from-port="0" to-layer="6704" to-port="1" />
-		<edge from-layer="6704" from-port="2" to-layer="6706" to-port="2" />
-		<edge from-layer="6705" from-port="0" to-layer="6706" to-port="3" />
-		<edge from-layer="6706" from-port="4" to-layer="6715" to-port="2" />
-		<edge from-layer="6706" from-port="4" to-layer="6709" to-port="1" />
-		<edge from-layer="6707" from-port="0" to-layer="6709" to-port="2" />
-		<edge from-layer="6708" from-port="0" to-layer="6709" to-port="3" />
-		<edge from-layer="6709" from-port="4" to-layer="6712" to-port="0" />
-		<edge from-layer="6710" from-port="0" to-layer="6711" to-port="0" />
-		<edge from-layer="6711" from-port="1" to-layer="6712" to-port="1" />
-		<edge from-layer="6712" from-port="2" to-layer="6716" to-port="0" />
-		<edge from-layer="6713" from-port="0" to-layer="6715" to-port="1" />
-		<edge from-layer="6714" from-port="0" to-layer="6715" to-port="3" />
-		<edge from-layer="6715" from-port="4" to-layer="6716" to-port="1" />
-		<edge from-layer="6716" from-port="2" to-layer="6725" to-port="0" />
-		<edge from-layer="6717" from-port="0" to-layer="6718" to-port="0" />
-		<edge from-layer="6718" from-port="1" to-layer="6721" to-port="0" />
-		<edge from-layer="6719" from-port="0" to-layer="6721" to-port="1" />
-		<edge from-layer="6720" from-port="0" to-layer="6721" to-port="3" />
-		<edge from-layer="6721" from-port="4" to-layer="6723" to-port="0" />
-		<edge from-layer="6722" from-port="0" to-layer="6723" to-port="2" />
-		<edge from-layer="6723" from-port="3" to-layer="6724" to-port="0" />
-		<edge from-layer="6724" from-port="2" to-layer="6725" to-port="1" />
-		<edge from-layer="6724" from-port="2" to-layer="6758" to-port="1" />
-		<edge from-layer="6725" from-port="2" to-layer="6726" to-port="1" />
-		<edge from-layer="6726" from-port="2" to-layer="6731" to-port="0" />
-		<edge from-layer="6727" from-port="0" to-layer="6730" to-port="1" />
-		<edge from-layer="6728" from-port="0" to-layer="6730" to-port="2" />
-		<edge from-layer="6729" from-port="0" to-layer="6730" to-port="3" />
-		<edge from-layer="6730" from-port="4" to-layer="6731" to-port="1" />
-		<edge from-layer="6731" from-port="2" to-layer="6789" to-port="0" />
-		<edge from-layer="6731" from-port="2" to-layer="6822" to-port="0" />
-		<edge from-layer="6732" from-port="0" to-layer="6735" to-port="1" />
-		<edge from-layer="6733" from-port="0" to-layer="6735" to-port="2" />
-		<edge from-layer="6734" from-port="0" to-layer="6735" to-port="3" />
-		<edge from-layer="6735" from-port="4" to-layer="6736" to-port="0" />
-		<edge from-layer="6735" from-port="4" to-layer="6739" to-port="0" />
-		<edge from-layer="6735" from-port="4" to-layer="6756" to-port="0" />
-		<edge from-layer="6735" from-port="4" to-layer="6750" to-port="0" />
-		<edge from-layer="6736" from-port="2" to-layer="6759" to-port="0" />
-		<edge from-layer="6737" from-port="0" to-layer="6747" to-port="0" />
-		<edge from-layer="6738" from-port="0" to-layer="6747" to-port="1" />
-		<edge from-layer="6739" from-port="1" to-layer="6741" to-port="0" />
-		<edge from-layer="6740" from-port="0" to-layer="6741" to-port="2" />
-		<edge from-layer="6741" from-port="3" to-layer="6742" to-port="0" />
-		<edge from-layer="6742" from-port="2" to-layer="6743" to-port="0" />
-		<edge from-layer="6743" from-port="1" to-layer="6745" to-port="0" />
-		<edge from-layer="6744" from-port="0" to-layer="6745" to-port="1" />
-		<edge from-layer="6745" from-port="2" to-layer="6747" to-port="2" />
-		<edge from-layer="6746" from-port="0" to-layer="6747" to-port="3" />
-		<edge from-layer="6747" from-port="4" to-layer="6750" to-port="1" />
-		<edge from-layer="6747" from-port="4" to-layer="6756" to-port="2" />
-		<edge from-layer="6748" from-port="0" to-layer="6750" to-port="2" />
-		<edge from-layer="6749" from-port="0" to-layer="6750" to-port="3" />
-		<edge from-layer="6750" from-port="4" to-layer="6753" to-port="0" />
-		<edge from-layer="6751" from-port="0" to-layer="6752" to-port="0" />
-		<edge from-layer="6752" from-port="1" to-layer="6753" to-port="1" />
-		<edge from-layer="6753" from-port="2" to-layer="6757" to-port="0" />
-		<edge from-layer="6754" from-port="0" to-layer="6756" to-port="1" />
-		<edge from-layer="6755" from-port="0" to-layer="6756" to-port="3" />
-		<edge from-layer="6756" from-port="4" to-layer="6757" to-port="1" />
-		<edge from-layer="6757" from-port="2" to-layer="6758" to-port="0" />
-		<edge from-layer="6758" from-port="2" to-layer="6759" to-port="1" />
-		<edge from-layer="6759" from-port="2" to-layer="6764" to-port="0" />
-		<edge from-layer="6760" from-port="0" to-layer="6763" to-port="1" />
-		<edge from-layer="6761" from-port="0" to-layer="6763" to-port="2" />
-		<edge from-layer="6762" from-port="0" to-layer="6763" to-port="3" />
-		<edge from-layer="6763" from-port="4" to-layer="6764" to-port="1" />
-		<edge from-layer="6764" from-port="2" to-layer="6765" to-port="1" />
-		<edge from-layer="6765" from-port="2" to-layer="6822" to-port="1" />
-		<edge from-layer="6765" from-port="2" to-layer="7682" to-port="0" />
-		<edge from-layer="6765" from-port="2" to-layer="6786" to-port="0" />
-		<edge from-layer="6766" from-port="0" to-layer="6767" to-port="0" />
-		<edge from-layer="6767" from-port="1" to-layer="6772" to-port="0" />
-		<edge from-layer="6768" from-port="0" to-layer="6771" to-port="1" />
-		<edge from-layer="6769" from-port="0" to-layer="6771" to-port="2" />
-		<edge from-layer="6770" from-port="0" to-layer="6771" to-port="3" />
-		<edge from-layer="6771" from-port="4" to-layer="6772" to-port="1" />
-		<edge from-layer="6772" from-port="2" to-layer="6773" to-port="0" />
-		<edge from-layer="6773" from-port="1" to-layer="6775" to-port="0" />
-		<edge from-layer="6774" from-port="0" to-layer="6775" to-port="2" />
-		<edge from-layer="6775" from-port="3" to-layer="6782" to-port="0" />
-		<edge from-layer="6776" from-port="0" to-layer="6779" to-port="1" />
-		<edge from-layer="6777" from-port="0" to-layer="6779" to-port="2" />
-		<edge from-layer="6778" from-port="0" to-layer="6779" to-port="3" />
-		<edge from-layer="6779" from-port="4" to-layer="6781" to-port="0" />
-		<edge from-layer="6780" from-port="0" to-layer="6781" to-port="1" />
-		<edge from-layer="6781" from-port="2" to-layer="6782" to-port="1" />
-		<edge from-layer="6782" from-port="2" to-layer="7681" to-port="0" />
-		<edge from-layer="6782" from-port="2" to-layer="6822" to-port="2" />
-		<edge from-layer="6783" from-port="0" to-layer="6801" to-port="0" />
-		<edge from-layer="6784" from-port="0" to-layer="6796" to-port="0" />
-		<edge from-layer="6785" from-port="0" to-layer="6799" to-port="1" />
-		<edge from-layer="6785" from-port="0" to-layer="6796" to-port="1" />
-		<edge from-layer="6786" from-port="1" to-layer="6788" to-port="0" />
-		<edge from-layer="6787" from-port="0" to-layer="6788" to-port="2" />
-		<edge from-layer="6788" from-port="3" to-layer="6792" to-port="0" />
-		<edge from-layer="6788" from-port="3" to-layer="6798" to-port="0" />
-		<edge from-layer="6789" from-port="1" to-layer="6791" to-port="0" />
-		<edge from-layer="6789" from-port="1" to-layer="6817" to-port="0" />
-		<edge from-layer="6790" from-port="0" to-layer="6791" to-port="2" />
-		<edge from-layer="6791" from-port="3" to-layer="6792" to-port="1" />
-		<edge from-layer="6792" from-port="2" to-layer="6794" to-port="0" />
-		<edge from-layer="6793" from-port="0" to-layer="6794" to-port="1" />
-		<edge from-layer="6793" from-port="0" to-layer="6798" to-port="1" />
-		<edge from-layer="6794" from-port="2" to-layer="6796" to-port="2" />
-		<edge from-layer="6795" from-port="0" to-layer="6796" to-port="3" />
-		<edge from-layer="6795" from-port="0" to-layer="6799" to-port="3" />
-		<edge from-layer="6796" from-port="4" to-layer="6801" to-port="1" />
-		<edge from-layer="6797" from-port="0" to-layer="6799" to-port="0" />
-		<edge from-layer="6798" from-port="2" to-layer="6799" to-port="2" />
-		<edge from-layer="6798" from-port="2" to-layer="6806" to-port="2" />
-		<edge from-layer="6799" from-port="4" to-layer="6801" to-port="2" />
-		<edge from-layer="6800" from-port="0" to-layer="6801" to-port="3" />
-		<edge from-layer="6801" from-port="4" to-layer="6808" to-port="0" />
-		<edge from-layer="6802" from-port="0" to-layer="6808" to-port="1" />
-		<edge from-layer="6803" from-port="0" to-layer="6806" to-port="0" />
-		<edge from-layer="6804" from-port="0" to-layer="6806" to-port="1" />
-		<edge from-layer="6805" from-port="0" to-layer="6806" to-port="3" />
-		<edge from-layer="6806" from-port="4" to-layer="6808" to-port="2" />
-		<edge from-layer="6807" from-port="0" to-layer="6808" to-port="3" />
-		<edge from-layer="6808" from-port="4" to-layer="6812" to-port="0" />
-		<edge from-layer="6809" from-port="0" to-layer="6810" to-port="0" />
-		<edge from-layer="6810" from-port="1" to-layer="6812" to-port="1" />
-		<edge from-layer="6811" from-port="0" to-layer="6812" to-port="2" />
-		<edge from-layer="6812" from-port="3" to-layer="6813" to-port="0" />
-		<edge from-layer="6813" from-port="2" to-layer="6814" to-port="0" />
-		<edge from-layer="6813" from-port="2" to-layer="6821" to-port="0" />
-		<edge from-layer="6814" from-port="1" to-layer="6820" to-port="0" />
-		<edge from-layer="6815" from-port="0" to-layer="6817" to-port="1" />
-		<edge from-layer="6816" from-port="0" to-layer="6817" to-port="2" />
-		<edge from-layer="6817" from-port="3" to-layer="6818" to-port="0" />
-		<edge from-layer="6818" from-port="4" to-layer="6819" to-port="0" />
-		<edge from-layer="6819" from-port="1" to-layer="6820" to-port="1" />
-		<edge from-layer="6820" from-port="2" to-layer="6821" to-port="1" />
-		<edge from-layer="6821" from-port="2" to-layer="6822" to-port="3" />
-		<edge from-layer="6822" from-port="4" to-layer="6824" to-port="0" />
-		<edge from-layer="6823" from-port="0" to-layer="6824" to-port="1" />
-		<edge from-layer="6824" from-port="2" to-layer="6826" to-port="0" />
-		<edge from-layer="6825" from-port="0" to-layer="6826" to-port="1" />
-		<edge from-layer="6826" from-port="2" to-layer="6835" to-port="0" />
-		<edge from-layer="6827" from-port="0" to-layer="6828" to-port="0" />
-		<edge from-layer="6828" from-port="1" to-layer="6831" to-port="0" />
-		<edge from-layer="6829" from-port="0" to-layer="6830" to-port="0" />
-		<edge from-layer="6830" from-port="1" to-layer="6831" to-port="1" />
-		<edge from-layer="6831" from-port="2" to-layer="6833" to-port="0" />
-		<edge from-layer="6832" from-port="0" to-layer="6833" to-port="1" />
-		<edge from-layer="6833" from-port="2" to-layer="6834" to-port="0" />
-		<edge from-layer="6834" from-port="1" to-layer="6835" to-port="1" />
-		<edge from-layer="6835" from-port="2" to-layer="6838" to-port="0" />
-		<edge from-layer="6836" from-port="0" to-layer="6837" to-port="0" />
-		<edge from-layer="6837" from-port="1" to-layer="6838" to-port="1" />
-		<edge from-layer="6838" from-port="2" to-layer="6839" to-port="1" />
-		<edge from-layer="6839" from-port="2" to-layer="6840" to-port="0" />
-		<edge from-layer="6840" from-port="2" to-layer="7082" to-port="1" />
-		<edge from-layer="6840" from-port="2" to-layer="6842" to-port="0" />
-		<edge from-layer="6841" from-port="0" to-layer="6842" to-port="1" />
-		<edge from-layer="6842" from-port="2" to-layer="6876" to-port="0" />
-		<edge from-layer="6842" from-port="2" to-layer="6845" to-port="0" />
-		<edge from-layer="6843" from-port="0" to-layer="6844" to-port="0" />
-		<edge from-layer="6844" from-port="1" to-layer="6845" to-port="1" />
-		<edge from-layer="6845" from-port="2" to-layer="6848" to-port="0" />
-		<edge from-layer="6846" from-port="0" to-layer="6847" to-port="0" />
-		<edge from-layer="6847" from-port="1" to-layer="6848" to-port="1" />
-		<edge from-layer="6848" from-port="2" to-layer="6857" to-port="0" />
-		<edge from-layer="6849" from-port="0" to-layer="6850" to-port="0" />
-		<edge from-layer="6850" from-port="1" to-layer="6853" to-port="0" />
-		<edge from-layer="6851" from-port="0" to-layer="6852" to-port="0" />
-		<edge from-layer="6852" from-port="1" to-layer="6853" to-port="1" />
-		<edge from-layer="6853" from-port="2" to-layer="6855" to-port="0" />
-		<edge from-layer="6854" from-port="0" to-layer="6855" to-port="1" />
-		<edge from-layer="6855" from-port="2" to-layer="6856" to-port="0" />
-		<edge from-layer="6856" from-port="1" to-layer="6857" to-port="1" />
-		<edge from-layer="6857" from-port="2" to-layer="6860" to-port="0" />
-		<edge from-layer="6858" from-port="0" to-layer="6859" to-port="0" />
-		<edge from-layer="6859" from-port="1" to-layer="6860" to-port="1" />
-		<edge from-layer="6860" from-port="2" to-layer="6861" to-port="0" />
-		<edge from-layer="6861" from-port="1" to-layer="6870" to-port="0" />
-		<edge from-layer="6862" from-port="0" to-layer="6863" to-port="0" />
-		<edge from-layer="6863" from-port="1" to-layer="6866" to-port="0" />
-		<edge from-layer="6864" from-port="0" to-layer="6865" to-port="0" />
-		<edge from-layer="6865" from-port="1" to-layer="6866" to-port="1" />
-		<edge from-layer="6866" from-port="2" to-layer="6868" to-port="0" />
-		<edge from-layer="6867" from-port="0" to-layer="6868" to-port="1" />
-		<edge from-layer="6868" from-port="2" to-layer="6869" to-port="0" />
-		<edge from-layer="6869" from-port="1" to-layer="6870" to-port="1" />
-		<edge from-layer="6870" from-port="2" to-layer="6873" to-port="0" />
-		<edge from-layer="6871" from-port="0" to-layer="6872" to-port="0" />
-		<edge from-layer="6872" from-port="1" to-layer="6873" to-port="1" />
-		<edge from-layer="6873" from-port="2" to-layer="7081" to-port="0" />
-		<edge from-layer="6874" from-port="0" to-layer="6875" to-port="0" />
-		<edge from-layer="6875" from-port="1" to-layer="6876" to-port="1" />
-		<edge from-layer="6876" from-port="2" to-layer="6879" to-port="0" />
-		<edge from-layer="6877" from-port="0" to-layer="6878" to-port="0" />
-		<edge from-layer="6878" from-port="1" to-layer="6879" to-port="1" />
-		<edge from-layer="6879" from-port="2" to-layer="6888" to-port="0" />
-		<edge from-layer="6880" from-port="0" to-layer="6881" to-port="0" />
-		<edge from-layer="6881" from-port="1" to-layer="6884" to-port="0" />
-		<edge from-layer="6882" from-port="0" to-layer="6883" to-port="0" />
-		<edge from-layer="6883" from-port="1" to-layer="6884" to-port="1" />
-		<edge from-layer="6884" from-port="2" to-layer="6886" to-port="0" />
-		<edge from-layer="6885" from-port="0" to-layer="6886" to-port="1" />
-		<edge from-layer="6886" from-port="2" to-layer="6887" to-port="0" />
-		<edge from-layer="6887" from-port="1" to-layer="6888" to-port="1" />
-		<edge from-layer="6888" from-port="2" to-layer="6891" to-port="0" />
-		<edge from-layer="6889" from-port="0" to-layer="6890" to-port="0" />
-		<edge from-layer="6890" from-port="1" to-layer="6891" to-port="1" />
-		<edge from-layer="6891" from-port="2" to-layer="6893" to-port="0" />
-		<edge from-layer="6892" from-port="0" to-layer="6893" to-port="1" />
-		<edge from-layer="6893" from-port="2" to-layer="7021" to-port="0" />
-		<edge from-layer="6893" from-port="2" to-layer="6897" to-port="0" />
-		<edge from-layer="6893" from-port="2" to-layer="6910" to-port="0" />
-		<edge from-layer="6894" from-port="0" to-layer="6897" to-port="1" />
-		<edge from-layer="6895" from-port="0" to-layer="6897" to-port="2" />
-		<edge from-layer="6896" from-port="0" to-layer="6897" to-port="3" />
-		<edge from-layer="6897" from-port="4" to-layer="6899" to-port="0" />
-		<edge from-layer="6898" from-port="0" to-layer="6899" to-port="1" />
-		<edge from-layer="6899" from-port="2" to-layer="6903" to-port="0" />
-		<edge from-layer="6899" from-port="2" to-layer="6972" to-port="0" />
-		<edge from-layer="6900" from-port="0" to-layer="6903" to-port="1" />
-		<edge from-layer="6901" from-port="0" to-layer="6903" to-port="2" />
-		<edge from-layer="6902" from-port="0" to-layer="6903" to-port="3" />
-		<edge from-layer="6903" from-port="4" to-layer="6951" to-port="0" />
-		<edge from-layer="6903" from-port="4" to-layer="6957" to-port="0" />
-		<edge from-layer="6903" from-port="4" to-layer="6937" to-port="0" />
-		<edge from-layer="6903" from-port="4" to-layer="6940" to-port="0" />
-		<edge from-layer="6904" from-port="0" to-layer="6905" to-port="0" />
-		<edge from-layer="6905" from-port="1" to-layer="6933" to-port="0" />
-		<edge from-layer="6906" from-port="0" to-layer="6933" to-port="1" />
-		<edge from-layer="6907" from-port="0" to-layer="6910" to-port="1" />
-		<edge from-layer="6908" from-port="0" to-layer="6910" to-port="2" />
-		<edge from-layer="6909" from-port="0" to-layer="6910" to-port="3" />
-		<edge from-layer="6910" from-port="4" to-layer="6912" to-port="0" />
-		<edge from-layer="6911" from-port="0" to-layer="6912" to-port="1" />
-		<edge from-layer="6912" from-port="2" to-layer="7005" to-port="0" />
-		<edge from-layer="6912" from-port="2" to-layer="6977" to-port="0" />
-		<edge from-layer="6912" from-port="2" to-layer="6913" to-port="0" />
-		<edge from-layer="6913" from-port="1" to-layer="6915" to-port="0" />
-		<edge from-layer="6914" from-port="0" to-layer="6915" to-port="2" />
-		<edge from-layer="6915" from-port="3" to-layer="6929" to-port="0" />
-		<edge from-layer="6916" from-port="0" to-layer="6917" to-port="0" />
-		<edge from-layer="6917" from-port="1" to-layer="6922" to-port="0" />
-		<edge from-layer="6918" from-port="0" to-layer="6921" to-port="1" />
-		<edge from-layer="6919" from-port="0" to-layer="6921" to-port="2" />
-		<edge from-layer="6920" from-port="0" to-layer="6921" to-port="3" />
-		<edge from-layer="6921" from-port="4" to-layer="6922" to-port="1" />
-		<edge from-layer="6922" from-port="2" to-layer="6923" to-port="0" />
-		<edge from-layer="6923" from-port="1" to-layer="6925" to-port="0" />
-		<edge from-layer="6924" from-port="0" to-layer="6925" to-port="2" />
-		<edge from-layer="6925" from-port="3" to-layer="6926" to-port="0" />
-		<edge from-layer="6925" from-port="3" to-layer="7007" to-port="0" />
-		<edge from-layer="6926" from-port="1" to-layer="6928" to-port="0" />
-		<edge from-layer="6927" from-port="0" to-layer="6928" to-port="2" />
-		<edge from-layer="6928" from-port="3" to-layer="6929" to-port="1" />
-		<edge from-layer="6929" from-port="2" to-layer="6931" to-port="0" />
-		<edge from-layer="6930" from-port="0" to-layer="6931" to-port="1" />
-		<edge from-layer="6931" from-port="2" to-layer="6933" to-port="2" />
-		<edge from-layer="6931" from-port="2" to-layer="6963" to-port="2" />
-		<edge from-layer="6932" from-port="0" to-layer="6933" to-port="3" />
-		<edge from-layer="6933" from-port="4" to-layer="6935" to-port="0" />
-		<edge from-layer="6934" from-port="0" to-layer="6935" to-port="2" />
-		<edge from-layer="6935" from-port="3" to-layer="6936" to-port="0" />
-		<edge from-layer="6936" from-port="2" to-layer="6937" to-port="1" />
-		<edge from-layer="6936" from-port="2" to-layer="6978" to-port="1" />
-		<edge from-layer="6937" from-port="2" to-layer="6968" to-port="0" />
-		<edge from-layer="6938" from-port="0" to-layer="6948" to-port="0" />
-		<edge from-layer="6939" from-port="0" to-layer="6948" to-port="1" />
-		<edge from-layer="6940" from-port="1" to-layer="6942" to-port="0" />
-		<edge from-layer="6941" from-port="0" to-layer="6942" to-port="2" />
-		<edge from-layer="6942" from-port="3" to-layer="6943" to-port="0" />
-		<edge from-layer="6943" from-port="2" to-layer="6944" to-port="0" />
-		<edge from-layer="6944" from-port="1" to-layer="6946" to-port="0" />
-		<edge from-layer="6945" from-port="0" to-layer="6946" to-port="1" />
-		<edge from-layer="6946" from-port="2" to-layer="6948" to-port="2" />
-		<edge from-layer="6947" from-port="0" to-layer="6948" to-port="3" />
-		<edge from-layer="6948" from-port="4" to-layer="6951" to-port="1" />
-		<edge from-layer="6948" from-port="4" to-layer="6957" to-port="2" />
-		<edge from-layer="6949" from-port="0" to-layer="6951" to-port="2" />
-		<edge from-layer="6950" from-port="0" to-layer="6951" to-port="3" />
-		<edge from-layer="6951" from-port="4" to-layer="6954" to-port="0" />
-		<edge from-layer="6952" from-port="0" to-layer="6953" to-port="0" />
-		<edge from-layer="6953" from-port="1" to-layer="6954" to-port="1" />
-		<edge from-layer="6954" from-port="2" to-layer="6958" to-port="0" />
-		<edge from-layer="6955" from-port="0" to-layer="6957" to-port="1" />
-		<edge from-layer="6956" from-port="0" to-layer="6957" to-port="3" />
-		<edge from-layer="6957" from-port="4" to-layer="6958" to-port="1" />
-		<edge from-layer="6958" from-port="2" to-layer="6967" to-port="0" />
-		<edge from-layer="6959" from-port="0" to-layer="6960" to-port="0" />
-		<edge from-layer="6960" from-port="1" to-layer="6963" to-port="0" />
-		<edge from-layer="6961" from-port="0" to-layer="6963" to-port="1" />
-		<edge from-layer="6962" from-port="0" to-layer="6963" to-port="3" />
-		<edge from-layer="6963" from-port="4" to-layer="6965" to-port="0" />
-		<edge from-layer="6964" from-port="0" to-layer="6965" to-port="2" />
-		<edge from-layer="6965" from-port="3" to-layer="6966" to-port="0" />
-		<edge from-layer="6966" from-port="2" to-layer="6967" to-port="1" />
-		<edge from-layer="6966" from-port="2" to-layer="7000" to-port="1" />
-		<edge from-layer="6967" from-port="2" to-layer="6968" to-port="1" />
-		<edge from-layer="6968" from-port="2" to-layer="6973" to-port="0" />
-		<edge from-layer="6969" from-port="0" to-layer="6972" to-port="1" />
-		<edge from-layer="6970" from-port="0" to-layer="6972" to-port="2" />
-		<edge from-layer="6971" from-port="0" to-layer="6972" to-port="3" />
-		<edge from-layer="6972" from-port="4" to-layer="6973" to-port="1" />
-		<edge from-layer="6973" from-port="2" to-layer="7064" to-port="0" />
-		<edge from-layer="6973" from-port="2" to-layer="7031" to-port="0" />
-		<edge from-layer="6974" from-port="0" to-layer="6977" to-port="1" />
-		<edge from-layer="6975" from-port="0" to-layer="6977" to-port="2" />
-		<edge from-layer="6976" from-port="0" to-layer="6977" to-port="3" />
-		<edge from-layer="6977" from-port="4" to-layer="6978" to-port="0" />
-		<edge from-layer="6977" from-port="4" to-layer="6998" to-port="0" />
-		<edge from-layer="6977" from-port="4" to-layer="6981" to-port="0" />
-		<edge from-layer="6977" from-port="4" to-layer="6992" to-port="0" />
-		<edge from-layer="6978" from-port="2" to-layer="7001" to-port="0" />
-		<edge from-layer="6979" from-port="0" to-layer="6989" to-port="0" />
-		<edge from-layer="6980" from-port="0" to-layer="6989" to-port="1" />
-		<edge from-layer="6981" from-port="1" to-layer="6983" to-port="0" />
-		<edge from-layer="6982" from-port="0" to-layer="6983" to-port="2" />
-		<edge from-layer="6983" from-port="3" to-layer="6984" to-port="0" />
-		<edge from-layer="6984" from-port="2" to-layer="6985" to-port="0" />
-		<edge from-layer="6985" from-port="1" to-layer="6987" to-port="0" />
-		<edge from-layer="6986" from-port="0" to-layer="6987" to-port="1" />
-		<edge from-layer="6987" from-port="2" to-layer="6989" to-port="2" />
-		<edge from-layer="6988" from-port="0" to-layer="6989" to-port="3" />
-		<edge from-layer="6989" from-port="4" to-layer="6998" to-port="2" />
-		<edge from-layer="6989" from-port="4" to-layer="6992" to-port="1" />
-		<edge from-layer="6990" from-port="0" to-layer="6992" to-port="2" />
-		<edge from-layer="6991" from-port="0" to-layer="6992" to-port="3" />
-		<edge from-layer="6992" from-port="4" to-layer="6995" to-port="0" />
-		<edge from-layer="6993" from-port="0" to-layer="6994" to-port="0" />
-		<edge from-layer="6994" from-port="1" to-layer="6995" to-port="1" />
-		<edge from-layer="6995" from-port="2" to-layer="6999" to-port="0" />
-		<edge from-layer="6996" from-port="0" to-layer="6998" to-port="1" />
-		<edge from-layer="6997" from-port="0" to-layer="6998" to-port="3" />
-		<edge from-layer="6998" from-port="4" to-layer="6999" to-port="1" />
-		<edge from-layer="6999" from-port="2" to-layer="7000" to-port="0" />
-		<edge from-layer="7000" from-port="2" to-layer="7001" to-port="1" />
-		<edge from-layer="7001" from-port="2" to-layer="7006" to-port="0" />
-		<edge from-layer="7002" from-port="0" to-layer="7005" to-port="1" />
-		<edge from-layer="7003" from-port="0" to-layer="7005" to-port="2" />
-		<edge from-layer="7004" from-port="0" to-layer="7005" to-port="3" />
-		<edge from-layer="7005" from-port="4" to-layer="7006" to-port="1" />
-		<edge from-layer="7006" from-port="2" to-layer="7007" to-port="1" />
-		<edge from-layer="7007" from-port="2" to-layer="7680" to-port="0" />
-		<edge from-layer="7007" from-port="2" to-layer="7064" to-port="1" />
-		<edge from-layer="7007" from-port="2" to-layer="7028" to-port="0" />
-		<edge from-layer="7008" from-port="0" to-layer="7009" to-port="0" />
-		<edge from-layer="7009" from-port="1" to-layer="7014" to-port="0" />
-		<edge from-layer="7010" from-port="0" to-layer="7013" to-port="1" />
-		<edge from-layer="7011" from-port="0" to-layer="7013" to-port="2" />
-		<edge from-layer="7012" from-port="0" to-layer="7013" to-port="3" />
-		<edge from-layer="7013" from-port="4" to-layer="7014" to-port="1" />
-		<edge from-layer="7014" from-port="2" to-layer="7015" to-port="0" />
-		<edge from-layer="7015" from-port="1" to-layer="7017" to-port="0" />
-		<edge from-layer="7016" from-port="0" to-layer="7017" to-port="2" />
-		<edge from-layer="7017" from-port="3" to-layer="7024" to-port="0" />
-		<edge from-layer="7018" from-port="0" to-layer="7021" to-port="1" />
-		<edge from-layer="7019" from-port="0" to-layer="7021" to-port="2" />
-		<edge from-layer="7020" from-port="0" to-layer="7021" to-port="3" />
-		<edge from-layer="7021" from-port="4" to-layer="7023" to-port="0" />
-		<edge from-layer="7022" from-port="0" to-layer="7023" to-port="1" />
-		<edge from-layer="7023" from-port="2" to-layer="7024" to-port="1" />
-		<edge from-layer="7024" from-port="2" to-layer="7064" to-port="2" />
-		<edge from-layer="7024" from-port="2" to-layer="7679" to-port="0" />
-		<edge from-layer="7025" from-port="0" to-layer="7043" to-port="0" />
-		<edge from-layer="7026" from-port="0" to-layer="7038" to-port="0" />
-		<edge from-layer="7027" from-port="0" to-layer="7041" to-port="1" />
-		<edge from-layer="7027" from-port="0" to-layer="7038" to-port="1" />
-		<edge from-layer="7028" from-port="1" to-layer="7030" to-port="0" />
-		<edge from-layer="7029" from-port="0" to-layer="7030" to-port="2" />
-		<edge from-layer="7030" from-port="3" to-layer="7034" to-port="0" />
-		<edge from-layer="7030" from-port="3" to-layer="7040" to-port="0" />
-		<edge from-layer="7031" from-port="1" to-layer="7033" to-port="0" />
-		<edge from-layer="7031" from-port="1" to-layer="7059" to-port="0" />
-		<edge from-layer="7032" from-port="0" to-layer="7033" to-port="2" />
-		<edge from-layer="7033" from-port="3" to-layer="7034" to-port="1" />
-		<edge from-layer="7034" from-port="2" to-layer="7036" to-port="0" />
-		<edge from-layer="7035" from-port="0" to-layer="7036" to-port="1" />
-		<edge from-layer="7035" from-port="0" to-layer="7040" to-port="1" />
-		<edge from-layer="7036" from-port="2" to-layer="7038" to-port="2" />
-		<edge from-layer="7037" from-port="0" to-layer="7038" to-port="3" />
-		<edge from-layer="7037" from-port="0" to-layer="7041" to-port="3" />
-		<edge from-layer="7038" from-port="4" to-layer="7043" to-port="1" />
-		<edge from-layer="7039" from-port="0" to-layer="7041" to-port="0" />
-		<edge from-layer="7040" from-port="2" to-layer="7048" to-port="2" />
-		<edge from-layer="7040" from-port="2" to-layer="7041" to-port="2" />
-		<edge from-layer="7041" from-port="4" to-layer="7043" to-port="2" />
-		<edge from-layer="7042" from-port="0" to-layer="7043" to-port="3" />
-		<edge from-layer="7043" from-port="4" to-layer="7050" to-port="0" />
-		<edge from-layer="7044" from-port="0" to-layer="7050" to-port="1" />
-		<edge from-layer="7045" from-port="0" to-layer="7048" to-port="0" />
-		<edge from-layer="7046" from-port="0" to-layer="7048" to-port="1" />
-		<edge from-layer="7047" from-port="0" to-layer="7048" to-port="3" />
-		<edge from-layer="7048" from-port="4" to-layer="7050" to-port="2" />
-		<edge from-layer="7049" from-port="0" to-layer="7050" to-port="3" />
-		<edge from-layer="7050" from-port="4" to-layer="7054" to-port="0" />
-		<edge from-layer="7051" from-port="0" to-layer="7052" to-port="0" />
-		<edge from-layer="7052" from-port="1" to-layer="7054" to-port="1" />
-		<edge from-layer="7053" from-port="0" to-layer="7054" to-port="2" />
-		<edge from-layer="7054" from-port="3" to-layer="7055" to-port="0" />
-		<edge from-layer="7055" from-port="2" to-layer="7056" to-port="0" />
-		<edge from-layer="7055" from-port="2" to-layer="7063" to-port="0" />
-		<edge from-layer="7056" from-port="1" to-layer="7062" to-port="0" />
-		<edge from-layer="7057" from-port="0" to-layer="7059" to-port="1" />
-		<edge from-layer="7058" from-port="0" to-layer="7059" to-port="2" />
-		<edge from-layer="7059" from-port="3" to-layer="7060" to-port="0" />
-		<edge from-layer="7060" from-port="4" to-layer="7061" to-port="0" />
-		<edge from-layer="7061" from-port="1" to-layer="7062" to-port="1" />
-		<edge from-layer="7062" from-port="2" to-layer="7063" to-port="1" />
-		<edge from-layer="7063" from-port="2" to-layer="7064" to-port="3" />
-		<edge from-layer="7064" from-port="4" to-layer="7066" to-port="0" />
-		<edge from-layer="7065" from-port="0" to-layer="7066" to-port="1" />
-		<edge from-layer="7066" from-port="2" to-layer="7068" to-port="0" />
-		<edge from-layer="7067" from-port="0" to-layer="7068" to-port="1" />
-		<edge from-layer="7068" from-port="2" to-layer="7077" to-port="0" />
-		<edge from-layer="7069" from-port="0" to-layer="7070" to-port="0" />
-		<edge from-layer="7070" from-port="1" to-layer="7073" to-port="0" />
-		<edge from-layer="7071" from-port="0" to-layer="7072" to-port="0" />
-		<edge from-layer="7072" from-port="1" to-layer="7073" to-port="1" />
-		<edge from-layer="7073" from-port="2" to-layer="7075" to-port="0" />
-		<edge from-layer="7074" from-port="0" to-layer="7075" to-port="1" />
-		<edge from-layer="7075" from-port="2" to-layer="7076" to-port="0" />
-		<edge from-layer="7076" from-port="1" to-layer="7077" to-port="1" />
-		<edge from-layer="7077" from-port="2" to-layer="7080" to-port="0" />
-		<edge from-layer="7078" from-port="0" to-layer="7079" to-port="0" />
-		<edge from-layer="7079" from-port="1" to-layer="7080" to-port="1" />
-		<edge from-layer="7080" from-port="2" to-layer="7081" to-port="1" />
-		<edge from-layer="7081" from-port="2" to-layer="7082" to-port="0" />
-		<edge from-layer="7082" from-port="2" to-layer="7324" to-port="1" />
-		<edge from-layer="7082" from-port="2" to-layer="7084" to-port="0" />
-		<edge from-layer="7083" from-port="0" to-layer="7084" to-port="1" />
-		<edge from-layer="7084" from-port="2" to-layer="7118" to-port="0" />
-		<edge from-layer="7084" from-port="2" to-layer="7087" to-port="0" />
-		<edge from-layer="7085" from-port="0" to-layer="7086" to-port="0" />
-		<edge from-layer="7086" from-port="1" to-layer="7087" to-port="1" />
-		<edge from-layer="7087" from-port="2" to-layer="7090" to-port="0" />
-		<edge from-layer="7088" from-port="0" to-layer="7089" to-port="0" />
-		<edge from-layer="7089" from-port="1" to-layer="7090" to-port="1" />
-		<edge from-layer="7090" from-port="2" to-layer="7099" to-port="0" />
-		<edge from-layer="7091" from-port="0" to-layer="7092" to-port="0" />
-		<edge from-layer="7092" from-port="1" to-layer="7095" to-port="0" />
-		<edge from-layer="7093" from-port="0" to-layer="7094" to-port="0" />
-		<edge from-layer="7094" from-port="1" to-layer="7095" to-port="1" />
-		<edge from-layer="7095" from-port="2" to-layer="7097" to-port="0" />
-		<edge from-layer="7096" from-port="0" to-layer="7097" to-port="1" />
-		<edge from-layer="7097" from-port="2" to-layer="7098" to-port="0" />
-		<edge from-layer="7098" from-port="1" to-layer="7099" to-port="1" />
-		<edge from-layer="7099" from-port="2" to-layer="7102" to-port="0" />
-		<edge from-layer="7100" from-port="0" to-layer="7101" to-port="0" />
-		<edge from-layer="7101" from-port="1" to-layer="7102" to-port="1" />
-		<edge from-layer="7102" from-port="2" to-layer="7103" to-port="0" />
-		<edge from-layer="7103" from-port="1" to-layer="7112" to-port="0" />
-		<edge from-layer="7104" from-port="0" to-layer="7105" to-port="0" />
-		<edge from-layer="7105" from-port="1" to-layer="7108" to-port="0" />
-		<edge from-layer="7106" from-port="0" to-layer="7107" to-port="0" />
-		<edge from-layer="7107" from-port="1" to-layer="7108" to-port="1" />
-		<edge from-layer="7108" from-port="2" to-layer="7110" to-port="0" />
-		<edge from-layer="7109" from-port="0" to-layer="7110" to-port="1" />
-		<edge from-layer="7110" from-port="2" to-layer="7111" to-port="0" />
-		<edge from-layer="7111" from-port="1" to-layer="7112" to-port="1" />
-		<edge from-layer="7112" from-port="2" to-layer="7115" to-port="0" />
-		<edge from-layer="7113" from-port="0" to-layer="7114" to-port="0" />
-		<edge from-layer="7114" from-port="1" to-layer="7115" to-port="1" />
-		<edge from-layer="7115" from-port="2" to-layer="7323" to-port="0" />
-		<edge from-layer="7116" from-port="0" to-layer="7117" to-port="0" />
-		<edge from-layer="7117" from-port="1" to-layer="7118" to-port="1" />
-		<edge from-layer="7118" from-port="2" to-layer="7121" to-port="0" />
-		<edge from-layer="7119" from-port="0" to-layer="7120" to-port="0" />
-		<edge from-layer="7120" from-port="1" to-layer="7121" to-port="1" />
-		<edge from-layer="7121" from-port="2" to-layer="7130" to-port="0" />
-		<edge from-layer="7122" from-port="0" to-layer="7123" to-port="0" />
-		<edge from-layer="7123" from-port="1" to-layer="7126" to-port="0" />
-		<edge from-layer="7124" from-port="0" to-layer="7125" to-port="0" />
-		<edge from-layer="7125" from-port="1" to-layer="7126" to-port="1" />
-		<edge from-layer="7126" from-port="2" to-layer="7128" to-port="0" />
-		<edge from-layer="7127" from-port="0" to-layer="7128" to-port="1" />
-		<edge from-layer="7128" from-port="2" to-layer="7129" to-port="0" />
-		<edge from-layer="7129" from-port="1" to-layer="7130" to-port="1" />
-		<edge from-layer="7130" from-port="2" to-layer="7133" to-port="0" />
-		<edge from-layer="7131" from-port="0" to-layer="7132" to-port="0" />
-		<edge from-layer="7132" from-port="1" to-layer="7133" to-port="1" />
-		<edge from-layer="7133" from-port="2" to-layer="7135" to-port="0" />
-		<edge from-layer="7134" from-port="0" to-layer="7135" to-port="1" />
-		<edge from-layer="7135" from-port="2" to-layer="7263" to-port="0" />
-		<edge from-layer="7135" from-port="2" to-layer="7152" to-port="0" />
-		<edge from-layer="7135" from-port="2" to-layer="7139" to-port="0" />
-		<edge from-layer="7136" from-port="0" to-layer="7139" to-port="1" />
-		<edge from-layer="7137" from-port="0" to-layer="7139" to-port="2" />
-		<edge from-layer="7138" from-port="0" to-layer="7139" to-port="3" />
-		<edge from-layer="7139" from-port="4" to-layer="7141" to-port="0" />
-		<edge from-layer="7140" from-port="0" to-layer="7141" to-port="1" />
-		<edge from-layer="7141" from-port="2" to-layer="7214" to-port="0" />
-		<edge from-layer="7141" from-port="2" to-layer="7145" to-port="0" />
-		<edge from-layer="7142" from-port="0" to-layer="7145" to-port="1" />
-		<edge from-layer="7143" from-port="0" to-layer="7145" to-port="2" />
-		<edge from-layer="7144" from-port="0" to-layer="7145" to-port="3" />
-		<edge from-layer="7145" from-port="4" to-layer="7199" to-port="0" />
-		<edge from-layer="7145" from-port="4" to-layer="7193" to-port="0" />
-		<edge from-layer="7145" from-port="4" to-layer="7182" to-port="0" />
-		<edge from-layer="7145" from-port="4" to-layer="7179" to-port="0" />
-		<edge from-layer="7146" from-port="0" to-layer="7147" to-port="0" />
-		<edge from-layer="7147" from-port="1" to-layer="7175" to-port="0" />
-		<edge from-layer="7148" from-port="0" to-layer="7175" to-port="1" />
-		<edge from-layer="7149" from-port="0" to-layer="7152" to-port="1" />
-		<edge from-layer="7150" from-port="0" to-layer="7152" to-port="2" />
-		<edge from-layer="7151" from-port="0" to-layer="7152" to-port="3" />
-		<edge from-layer="7152" from-port="4" to-layer="7154" to-port="0" />
-		<edge from-layer="7153" from-port="0" to-layer="7154" to-port="1" />
-		<edge from-layer="7154" from-port="2" to-layer="7219" to-port="0" />
-		<edge from-layer="7154" from-port="2" to-layer="7155" to-port="0" />
-		<edge from-layer="7154" from-port="2" to-layer="7247" to-port="0" />
-		<edge from-layer="7155" from-port="1" to-layer="7157" to-port="0" />
-		<edge from-layer="7156" from-port="0" to-layer="7157" to-port="2" />
-		<edge from-layer="7157" from-port="3" to-layer="7171" to-port="0" />
-		<edge from-layer="7158" from-port="0" to-layer="7159" to-port="0" />
-		<edge from-layer="7159" from-port="1" to-layer="7164" to-port="0" />
-		<edge from-layer="7160" from-port="0" to-layer="7163" to-port="1" />
-		<edge from-layer="7161" from-port="0" to-layer="7163" to-port="2" />
-		<edge from-layer="7162" from-port="0" to-layer="7163" to-port="3" />
-		<edge from-layer="7163" from-port="4" to-layer="7164" to-port="1" />
-		<edge from-layer="7164" from-port="2" to-layer="7165" to-port="0" />
-		<edge from-layer="7165" from-port="1" to-layer="7167" to-port="0" />
-		<edge from-layer="7166" from-port="0" to-layer="7167" to-port="2" />
-		<edge from-layer="7167" from-port="3" to-layer="7168" to-port="0" />
-		<edge from-layer="7167" from-port="3" to-layer="7249" to-port="0" />
-		<edge from-layer="7168" from-port="1" to-layer="7170" to-port="0" />
-		<edge from-layer="7169" from-port="0" to-layer="7170" to-port="2" />
-		<edge from-layer="7170" from-port="3" to-layer="7171" to-port="1" />
-		<edge from-layer="7171" from-port="2" to-layer="7173" to-port="0" />
-		<edge from-layer="7172" from-port="0" to-layer="7173" to-port="1" />
-		<edge from-layer="7173" from-port="2" to-layer="7175" to-port="2" />
-		<edge from-layer="7173" from-port="2" to-layer="7205" to-port="2" />
-		<edge from-layer="7174" from-port="0" to-layer="7175" to-port="3" />
-		<edge from-layer="7175" from-port="4" to-layer="7177" to-port="0" />
-		<edge from-layer="7176" from-port="0" to-layer="7177" to-port="2" />
-		<edge from-layer="7177" from-port="3" to-layer="7178" to-port="0" />
-		<edge from-layer="7178" from-port="2" to-layer="7179" to-port="1" />
-		<edge from-layer="7178" from-port="2" to-layer="7220" to-port="1" />
-		<edge from-layer="7179" from-port="2" to-layer="7210" to-port="0" />
-		<edge from-layer="7180" from-port="0" to-layer="7190" to-port="0" />
-		<edge from-layer="7181" from-port="0" to-layer="7190" to-port="1" />
-		<edge from-layer="7182" from-port="1" to-layer="7184" to-port="0" />
-		<edge from-layer="7183" from-port="0" to-layer="7184" to-port="2" />
-		<edge from-layer="7184" from-port="3" to-layer="7185" to-port="0" />
-		<edge from-layer="7185" from-port="2" to-layer="7186" to-port="0" />
-		<edge from-layer="7186" from-port="1" to-layer="7188" to-port="0" />
-		<edge from-layer="7187" from-port="0" to-layer="7188" to-port="1" />
-		<edge from-layer="7188" from-port="2" to-layer="7190" to-port="2" />
-		<edge from-layer="7189" from-port="0" to-layer="7190" to-port="3" />
-		<edge from-layer="7190" from-port="4" to-layer="7193" to-port="1" />
-		<edge from-layer="7190" from-port="4" to-layer="7199" to-port="2" />
-		<edge from-layer="7191" from-port="0" to-layer="7193" to-port="2" />
-		<edge from-layer="7192" from-port="0" to-layer="7193" to-port="3" />
-		<edge from-layer="7193" from-port="4" to-layer="7196" to-port="0" />
-		<edge from-layer="7194" from-port="0" to-layer="7195" to-port="0" />
-		<edge from-layer="7195" from-port="1" to-layer="7196" to-port="1" />
-		<edge from-layer="7196" from-port="2" to-layer="7200" to-port="0" />
-		<edge from-layer="7197" from-port="0" to-layer="7199" to-port="1" />
-		<edge from-layer="7198" from-port="0" to-layer="7199" to-port="3" />
-		<edge from-layer="7199" from-port="4" to-layer="7200" to-port="1" />
-		<edge from-layer="7200" from-port="2" to-layer="7209" to-port="0" />
-		<edge from-layer="7201" from-port="0" to-layer="7202" to-port="0" />
-		<edge from-layer="7202" from-port="1" to-layer="7205" to-port="0" />
-		<edge from-layer="7203" from-port="0" to-layer="7205" to-port="1" />
-		<edge from-layer="7204" from-port="0" to-layer="7205" to-port="3" />
-		<edge from-layer="7205" from-port="4" to-layer="7207" to-port="0" />
-		<edge from-layer="7206" from-port="0" to-layer="7207" to-port="2" />
-		<edge from-layer="7207" from-port="3" to-layer="7208" to-port="0" />
-		<edge from-layer="7208" from-port="2" to-layer="7209" to-port="1" />
-		<edge from-layer="7208" from-port="2" to-layer="7242" to-port="1" />
-		<edge from-layer="7209" from-port="2" to-layer="7210" to-port="1" />
-		<edge from-layer="7210" from-port="2" to-layer="7215" to-port="0" />
-		<edge from-layer="7211" from-port="0" to-layer="7214" to-port="1" />
-		<edge from-layer="7212" from-port="0" to-layer="7214" to-port="2" />
-		<edge from-layer="7213" from-port="0" to-layer="7214" to-port="3" />
-		<edge from-layer="7214" from-port="4" to-layer="7215" to-port="1" />
-		<edge from-layer="7215" from-port="2" to-layer="7306" to-port="0" />
-		<edge from-layer="7215" from-port="2" to-layer="7273" to-port="0" />
-		<edge from-layer="7216" from-port="0" to-layer="7219" to-port="1" />
-		<edge from-layer="7217" from-port="0" to-layer="7219" to-port="2" />
-		<edge from-layer="7218" from-port="0" to-layer="7219" to-port="3" />
-		<edge from-layer="7219" from-port="4" to-layer="7240" to-port="0" />
-		<edge from-layer="7219" from-port="4" to-layer="7223" to-port="0" />
-		<edge from-layer="7219" from-port="4" to-layer="7234" to-port="0" />
-		<edge from-layer="7219" from-port="4" to-layer="7220" to-port="0" />
-		<edge from-layer="7220" from-port="2" to-layer="7243" to-port="0" />
-		<edge from-layer="7221" from-port="0" to-layer="7231" to-port="0" />
-		<edge from-layer="7222" from-port="0" to-layer="7231" to-port="1" />
-		<edge from-layer="7223" from-port="1" to-layer="7225" to-port="0" />
-		<edge from-layer="7224" from-port="0" to-layer="7225" to-port="2" />
-		<edge from-layer="7225" from-port="3" to-layer="7226" to-port="0" />
-		<edge from-layer="7226" from-port="2" to-layer="7227" to-port="0" />
-		<edge from-layer="7227" from-port="1" to-layer="7229" to-port="0" />
-		<edge from-layer="7228" from-port="0" to-layer="7229" to-port="1" />
-		<edge from-layer="7229" from-port="2" to-layer="7231" to-port="2" />
-		<edge from-layer="7230" from-port="0" to-layer="7231" to-port="3" />
-		<edge from-layer="7231" from-port="4" to-layer="7240" to-port="2" />
-		<edge from-layer="7231" from-port="4" to-layer="7234" to-port="1" />
-		<edge from-layer="7232" from-port="0" to-layer="7234" to-port="2" />
-		<edge from-layer="7233" from-port="0" to-layer="7234" to-port="3" />
-		<edge from-layer="7234" from-port="4" to-layer="7237" to-port="0" />
-		<edge from-layer="7235" from-port="0" to-layer="7236" to-port="0" />
-		<edge from-layer="7236" from-port="1" to-layer="7237" to-port="1" />
-		<edge from-layer="7237" from-port="2" to-layer="7241" to-port="0" />
-		<edge from-layer="7238" from-port="0" to-layer="7240" to-port="1" />
-		<edge from-layer="7239" from-port="0" to-layer="7240" to-port="3" />
-		<edge from-layer="7240" from-port="4" to-layer="7241" to-port="1" />
-		<edge from-layer="7241" from-port="2" to-layer="7242" to-port="0" />
-		<edge from-layer="7242" from-port="2" to-layer="7243" to-port="1" />
-		<edge from-layer="7243" from-port="2" to-layer="7248" to-port="0" />
-		<edge from-layer="7244" from-port="0" to-layer="7247" to-port="1" />
-		<edge from-layer="7245" from-port="0" to-layer="7247" to-port="2" />
-		<edge from-layer="7246" from-port="0" to-layer="7247" to-port="3" />
-		<edge from-layer="7247" from-port="4" to-layer="7248" to-port="1" />
-		<edge from-layer="7248" from-port="2" to-layer="7249" to-port="1" />
-		<edge from-layer="7249" from-port="2" to-layer="7678" to-port="0" />
-		<edge from-layer="7249" from-port="2" to-layer="7306" to-port="1" />
-		<edge from-layer="7249" from-port="2" to-layer="7270" to-port="0" />
-		<edge from-layer="7250" from-port="0" to-layer="7251" to-port="0" />
-		<edge from-layer="7251" from-port="1" to-layer="7256" to-port="0" />
-		<edge from-layer="7252" from-port="0" to-layer="7255" to-port="1" />
-		<edge from-layer="7253" from-port="0" to-layer="7255" to-port="2" />
-		<edge from-layer="7254" from-port="0" to-layer="7255" to-port="3" />
-		<edge from-layer="7255" from-port="4" to-layer="7256" to-port="1" />
-		<edge from-layer="7256" from-port="2" to-layer="7257" to-port="0" />
-		<edge from-layer="7257" from-port="1" to-layer="7259" to-port="0" />
-		<edge from-layer="7258" from-port="0" to-layer="7259" to-port="2" />
-		<edge from-layer="7259" from-port="3" to-layer="7266" to-port="0" />
-		<edge from-layer="7260" from-port="0" to-layer="7263" to-port="1" />
-		<edge from-layer="7261" from-port="0" to-layer="7263" to-port="2" />
-		<edge from-layer="7262" from-port="0" to-layer="7263" to-port="3" />
-		<edge from-layer="7263" from-port="4" to-layer="7265" to-port="0" />
-		<edge from-layer="7264" from-port="0" to-layer="7265" to-port="1" />
-		<edge from-layer="7265" from-port="2" to-layer="7266" to-port="1" />
-		<edge from-layer="7266" from-port="2" to-layer="7677" to-port="0" />
-		<edge from-layer="7266" from-port="2" to-layer="7306" to-port="2" />
-		<edge from-layer="7267" from-port="0" to-layer="7285" to-port="0" />
-		<edge from-layer="7268" from-port="0" to-layer="7280" to-port="0" />
-		<edge from-layer="7269" from-port="0" to-layer="7280" to-port="1" />
-		<edge from-layer="7269" from-port="0" to-layer="7283" to-port="1" />
-		<edge from-layer="7270" from-port="1" to-layer="7272" to-port="0" />
-		<edge from-layer="7271" from-port="0" to-layer="7272" to-port="2" />
-		<edge from-layer="7272" from-port="3" to-layer="7276" to-port="0" />
-		<edge from-layer="7272" from-port="3" to-layer="7282" to-port="0" />
-		<edge from-layer="7273" from-port="1" to-layer="7275" to-port="0" />
-		<edge from-layer="7273" from-port="1" to-layer="7301" to-port="0" />
-		<edge from-layer="7274" from-port="0" to-layer="7275" to-port="2" />
-		<edge from-layer="7275" from-port="3" to-layer="7276" to-port="1" />
-		<edge from-layer="7276" from-port="2" to-layer="7278" to-port="0" />
-		<edge from-layer="7277" from-port="0" to-layer="7278" to-port="1" />
-		<edge from-layer="7277" from-port="0" to-layer="7282" to-port="1" />
-		<edge from-layer="7278" from-port="2" to-layer="7280" to-port="2" />
-		<edge from-layer="7279" from-port="0" to-layer="7283" to-port="3" />
-		<edge from-layer="7279" from-port="0" to-layer="7280" to-port="3" />
-		<edge from-layer="7280" from-port="4" to-layer="7285" to-port="1" />
-		<edge from-layer="7281" from-port="0" to-layer="7283" to-port="0" />
-		<edge from-layer="7282" from-port="2" to-layer="7290" to-port="2" />
-		<edge from-layer="7282" from-port="2" to-layer="7283" to-port="2" />
-		<edge from-layer="7283" from-port="4" to-layer="7285" to-port="2" />
-		<edge from-layer="7284" from-port="0" to-layer="7285" to-port="3" />
-		<edge from-layer="7285" from-port="4" to-layer="7292" to-port="0" />
-		<edge from-layer="7286" from-port="0" to-layer="7292" to-port="1" />
-		<edge from-layer="7287" from-port="0" to-layer="7290" to-port="0" />
-		<edge from-layer="7288" from-port="0" to-layer="7290" to-port="1" />
-		<edge from-layer="7289" from-port="0" to-layer="7290" to-port="3" />
-		<edge from-layer="7290" from-port="4" to-layer="7292" to-port="2" />
-		<edge from-layer="7291" from-port="0" to-layer="7292" to-port="3" />
-		<edge from-layer="7292" from-port="4" to-layer="7296" to-port="0" />
-		<edge from-layer="7293" from-port="0" to-layer="7294" to-port="0" />
-		<edge from-layer="7294" from-port="1" to-layer="7296" to-port="1" />
-		<edge from-layer="7295" from-port="0" to-layer="7296" to-port="2" />
-		<edge from-layer="7296" from-port="3" to-layer="7297" to-port="0" />
-		<edge from-layer="7297" from-port="2" to-layer="7298" to-port="0" />
-		<edge from-layer="7297" from-port="2" to-layer="7305" to-port="0" />
-		<edge from-layer="7298" from-port="1" to-layer="7304" to-port="0" />
-		<edge from-layer="7299" from-port="0" to-layer="7301" to-port="1" />
-		<edge from-layer="7300" from-port="0" to-layer="7301" to-port="2" />
-		<edge from-layer="7301" from-port="3" to-layer="7302" to-port="0" />
-		<edge from-layer="7302" from-port="4" to-layer="7303" to-port="0" />
-		<edge from-layer="7303" from-port="1" to-layer="7304" to-port="1" />
-		<edge from-layer="7304" from-port="2" to-layer="7305" to-port="1" />
-		<edge from-layer="7305" from-port="2" to-layer="7306" to-port="3" />
-		<edge from-layer="7306" from-port="4" to-layer="7308" to-port="0" />
-		<edge from-layer="7307" from-port="0" to-layer="7308" to-port="1" />
-		<edge from-layer="7308" from-port="2" to-layer="7310" to-port="0" />
-		<edge from-layer="7309" from-port="0" to-layer="7310" to-port="1" />
-		<edge from-layer="7310" from-port="2" to-layer="7319" to-port="0" />
-		<edge from-layer="7311" from-port="0" to-layer="7312" to-port="0" />
-		<edge from-layer="7312" from-port="1" to-layer="7315" to-port="0" />
-		<edge from-layer="7313" from-port="0" to-layer="7314" to-port="0" />
-		<edge from-layer="7314" from-port="1" to-layer="7315" to-port="1" />
-		<edge from-layer="7315" from-port="2" to-layer="7317" to-port="0" />
-		<edge from-layer="7316" from-port="0" to-layer="7317" to-port="1" />
-		<edge from-layer="7317" from-port="2" to-layer="7318" to-port="0" />
-		<edge from-layer="7318" from-port="1" to-layer="7319" to-port="1" />
-		<edge from-layer="7319" from-port="2" to-layer="7322" to-port="0" />
-		<edge from-layer="7320" from-port="0" to-layer="7321" to-port="0" />
-		<edge from-layer="7321" from-port="1" to-layer="7322" to-port="1" />
-		<edge from-layer="7322" from-port="2" to-layer="7323" to-port="1" />
-		<edge from-layer="7323" from-port="2" to-layer="7324" to-port="0" />
-		<edge from-layer="7324" from-port="2" to-layer="7566" to-port="1" />
-		<edge from-layer="7324" from-port="2" to-layer="7326" to-port="0" />
-		<edge from-layer="7325" from-port="0" to-layer="7326" to-port="1" />
-		<edge from-layer="7326" from-port="2" to-layer="7329" to-port="0" />
-		<edge from-layer="7326" from-port="2" to-layer="7360" to-port="0" />
-		<edge from-layer="7327" from-port="0" to-layer="7328" to-port="0" />
-		<edge from-layer="7328" from-port="1" to-layer="7329" to-port="1" />
-		<edge from-layer="7329" from-port="2" to-layer="7332" to-port="0" />
-		<edge from-layer="7330" from-port="0" to-layer="7331" to-port="0" />
-		<edge from-layer="7331" from-port="1" to-layer="7332" to-port="1" />
-		<edge from-layer="7332" from-port="2" to-layer="7341" to-port="0" />
-		<edge from-layer="7333" from-port="0" to-layer="7334" to-port="0" />
-		<edge from-layer="7334" from-port="1" to-layer="7337" to-port="0" />
-		<edge from-layer="7335" from-port="0" to-layer="7336" to-port="0" />
-		<edge from-layer="7336" from-port="1" to-layer="7337" to-port="1" />
-		<edge from-layer="7337" from-port="2" to-layer="7339" to-port="0" />
-		<edge from-layer="7338" from-port="0" to-layer="7339" to-port="1" />
-		<edge from-layer="7339" from-port="2" to-layer="7340" to-port="0" />
-		<edge from-layer="7340" from-port="1" to-layer="7341" to-port="1" />
-		<edge from-layer="7341" from-port="2" to-layer="7344" to-port="0" />
-		<edge from-layer="7342" from-port="0" to-layer="7343" to-port="0" />
-		<edge from-layer="7343" from-port="1" to-layer="7344" to-port="1" />
-		<edge from-layer="7344" from-port="2" to-layer="7345" to-port="0" />
-		<edge from-layer="7345" from-port="1" to-layer="7354" to-port="0" />
-		<edge from-layer="7346" from-port="0" to-layer="7347" to-port="0" />
-		<edge from-layer="7347" from-port="1" to-layer="7350" to-port="0" />
-		<edge from-layer="7348" from-port="0" to-layer="7349" to-port="0" />
-		<edge from-layer="7349" from-port="1" to-layer="7350" to-port="1" />
-		<edge from-layer="7350" from-port="2" to-layer="7352" to-port="0" />
-		<edge from-layer="7351" from-port="0" to-layer="7352" to-port="1" />
-		<edge from-layer="7352" from-port="2" to-layer="7353" to-port="0" />
-		<edge from-layer="7353" from-port="1" to-layer="7354" to-port="1" />
-		<edge from-layer="7354" from-port="2" to-layer="7357" to-port="0" />
-		<edge from-layer="7355" from-port="0" to-layer="7356" to-port="0" />
-		<edge from-layer="7356" from-port="1" to-layer="7357" to-port="1" />
-		<edge from-layer="7357" from-port="2" to-layer="7565" to-port="0" />
-		<edge from-layer="7358" from-port="0" to-layer="7359" to-port="0" />
-		<edge from-layer="7359" from-port="1" to-layer="7360" to-port="1" />
-		<edge from-layer="7360" from-port="2" to-layer="7363" to-port="0" />
-		<edge from-layer="7361" from-port="0" to-layer="7362" to-port="0" />
-		<edge from-layer="7362" from-port="1" to-layer="7363" to-port="1" />
-		<edge from-layer="7363" from-port="2" to-layer="7372" to-port="0" />
-		<edge from-layer="7364" from-port="0" to-layer="7365" to-port="0" />
-		<edge from-layer="7365" from-port="1" to-layer="7368" to-port="0" />
-		<edge from-layer="7366" from-port="0" to-layer="7367" to-port="0" />
-		<edge from-layer="7367" from-port="1" to-layer="7368" to-port="1" />
-		<edge from-layer="7368" from-port="2" to-layer="7370" to-port="0" />
-		<edge from-layer="7369" from-port="0" to-layer="7370" to-port="1" />
-		<edge from-layer="7370" from-port="2" to-layer="7371" to-port="0" />
-		<edge from-layer="7371" from-port="1" to-layer="7372" to-port="1" />
-		<edge from-layer="7372" from-port="2" to-layer="7375" to-port="0" />
-		<edge from-layer="7373" from-port="0" to-layer="7374" to-port="0" />
-		<edge from-layer="7374" from-port="1" to-layer="7375" to-port="1" />
-		<edge from-layer="7375" from-port="2" to-layer="7377" to-port="0" />
-		<edge from-layer="7376" from-port="0" to-layer="7377" to-port="1" />
-		<edge from-layer="7377" from-port="2" to-layer="7394" to-port="0" />
-		<edge from-layer="7377" from-port="2" to-layer="7381" to-port="0" />
-		<edge from-layer="7377" from-port="2" to-layer="7505" to-port="0" />
-		<edge from-layer="7378" from-port="0" to-layer="7381" to-port="1" />
-		<edge from-layer="7379" from-port="0" to-layer="7381" to-port="2" />
-		<edge from-layer="7380" from-port="0" to-layer="7381" to-port="3" />
-		<edge from-layer="7381" from-port="4" to-layer="7383" to-port="0" />
-		<edge from-layer="7382" from-port="0" to-layer="7383" to-port="1" />
-		<edge from-layer="7383" from-port="2" to-layer="7456" to-port="0" />
-		<edge from-layer="7383" from-port="2" to-layer="7387" to-port="0" />
-		<edge from-layer="7384" from-port="0" to-layer="7387" to-port="1" />
-		<edge from-layer="7385" from-port="0" to-layer="7387" to-port="2" />
-		<edge from-layer="7386" from-port="0" to-layer="7387" to-port="3" />
-		<edge from-layer="7387" from-port="4" to-layer="7424" to-port="0" />
-		<edge from-layer="7387" from-port="4" to-layer="7421" to-port="0" />
-		<edge from-layer="7387" from-port="4" to-layer="7441" to-port="0" />
-		<edge from-layer="7387" from-port="4" to-layer="7435" to-port="0" />
-		<edge from-layer="7388" from-port="0" to-layer="7389" to-port="0" />
-		<edge from-layer="7389" from-port="1" to-layer="7417" to-port="0" />
-		<edge from-layer="7390" from-port="0" to-layer="7417" to-port="1" />
-		<edge from-layer="7391" from-port="0" to-layer="7394" to-port="1" />
-		<edge from-layer="7392" from-port="0" to-layer="7394" to-port="2" />
-		<edge from-layer="7393" from-port="0" to-layer="7394" to-port="3" />
-		<edge from-layer="7394" from-port="4" to-layer="7396" to-port="0" />
-		<edge from-layer="7395" from-port="0" to-layer="7396" to-port="1" />
-		<edge from-layer="7396" from-port="2" to-layer="7397" to-port="0" />
-		<edge from-layer="7396" from-port="2" to-layer="7489" to-port="0" />
-		<edge from-layer="7396" from-port="2" to-layer="7461" to-port="0" />
-		<edge from-layer="7397" from-port="1" to-layer="7399" to-port="0" />
-		<edge from-layer="7398" from-port="0" to-layer="7399" to-port="2" />
-		<edge from-layer="7399" from-port="3" to-layer="7413" to-port="0" />
-		<edge from-layer="7400" from-port="0" to-layer="7401" to-port="0" />
-		<edge from-layer="7401" from-port="1" to-layer="7406" to-port="0" />
-		<edge from-layer="7402" from-port="0" to-layer="7405" to-port="1" />
-		<edge from-layer="7403" from-port="0" to-layer="7405" to-port="2" />
-		<edge from-layer="7404" from-port="0" to-layer="7405" to-port="3" />
-		<edge from-layer="7405" from-port="4" to-layer="7406" to-port="1" />
-		<edge from-layer="7406" from-port="2" to-layer="7407" to-port="0" />
-		<edge from-layer="7407" from-port="1" to-layer="7409" to-port="0" />
-		<edge from-layer="7408" from-port="0" to-layer="7409" to-port="2" />
-		<edge from-layer="7409" from-port="3" to-layer="7410" to-port="0" />
-		<edge from-layer="7409" from-port="3" to-layer="7491" to-port="0" />
-		<edge from-layer="7410" from-port="1" to-layer="7412" to-port="0" />
-		<edge from-layer="7411" from-port="0" to-layer="7412" to-port="2" />
-		<edge from-layer="7412" from-port="3" to-layer="7413" to-port="1" />
-		<edge from-layer="7413" from-port="2" to-layer="7415" to-port="0" />
-		<edge from-layer="7414" from-port="0" to-layer="7415" to-port="1" />
-		<edge from-layer="7415" from-port="2" to-layer="7447" to-port="2" />
-		<edge from-layer="7415" from-port="2" to-layer="7417" to-port="2" />
-		<edge from-layer="7416" from-port="0" to-layer="7417" to-port="3" />
-		<edge from-layer="7417" from-port="4" to-layer="7419" to-port="0" />
-		<edge from-layer="7418" from-port="0" to-layer="7419" to-port="2" />
-		<edge from-layer="7419" from-port="3" to-layer="7420" to-port="0" />
-		<edge from-layer="7420" from-port="2" to-layer="7462" to-port="1" />
-		<edge from-layer="7420" from-port="2" to-layer="7421" to-port="1" />
-		<edge from-layer="7421" from-port="2" to-layer="7452" to-port="0" />
-		<edge from-layer="7422" from-port="0" to-layer="7432" to-port="0" />
-		<edge from-layer="7423" from-port="0" to-layer="7432" to-port="1" />
-		<edge from-layer="7424" from-port="1" to-layer="7426" to-port="0" />
-		<edge from-layer="7425" from-port="0" to-layer="7426" to-port="2" />
-		<edge from-layer="7426" from-port="3" to-layer="7427" to-port="0" />
-		<edge from-layer="7427" from-port="2" to-layer="7428" to-port="0" />
-		<edge from-layer="7428" from-port="1" to-layer="7430" to-port="0" />
-		<edge from-layer="7429" from-port="0" to-layer="7430" to-port="1" />
-		<edge from-layer="7430" from-port="2" to-layer="7432" to-port="2" />
-		<edge from-layer="7431" from-port="0" to-layer="7432" to-port="3" />
-		<edge from-layer="7432" from-port="4" to-layer="7441" to-port="2" />
-		<edge from-layer="7432" from-port="4" to-layer="7435" to-port="1" />
-		<edge from-layer="7433" from-port="0" to-layer="7435" to-port="2" />
-		<edge from-layer="7434" from-port="0" to-layer="7435" to-port="3" />
-		<edge from-layer="7435" from-port="4" to-layer="7438" to-port="0" />
-		<edge from-layer="7436" from-port="0" to-layer="7437" to-port="0" />
-		<edge from-layer="7437" from-port="1" to-layer="7438" to-port="1" />
-		<edge from-layer="7438" from-port="2" to-layer="7442" to-port="0" />
-		<edge from-layer="7439" from-port="0" to-layer="7441" to-port="1" />
-		<edge from-layer="7440" from-port="0" to-layer="7441" to-port="3" />
-		<edge from-layer="7441" from-port="4" to-layer="7442" to-port="1" />
-		<edge from-layer="7442" from-port="2" to-layer="7451" to-port="0" />
-		<edge from-layer="7443" from-port="0" to-layer="7444" to-port="0" />
-		<edge from-layer="7444" from-port="1" to-layer="7447" to-port="0" />
-		<edge from-layer="7445" from-port="0" to-layer="7447" to-port="1" />
-		<edge from-layer="7446" from-port="0" to-layer="7447" to-port="3" />
-		<edge from-layer="7447" from-port="4" to-layer="7449" to-port="0" />
-		<edge from-layer="7448" from-port="0" to-layer="7449" to-port="2" />
-		<edge from-layer="7449" from-port="3" to-layer="7450" to-port="0" />
-		<edge from-layer="7450" from-port="2" to-layer="7484" to-port="1" />
-		<edge from-layer="7450" from-port="2" to-layer="7451" to-port="1" />
-		<edge from-layer="7451" from-port="2" to-layer="7452" to-port="1" />
-		<edge from-layer="7452" from-port="2" to-layer="7457" to-port="0" />
-		<edge from-layer="7453" from-port="0" to-layer="7456" to-port="1" />
-		<edge from-layer="7454" from-port="0" to-layer="7456" to-port="2" />
-		<edge from-layer="7455" from-port="0" to-layer="7456" to-port="3" />
-		<edge from-layer="7456" from-port="4" to-layer="7457" to-port="1" />
-		<edge from-layer="7457" from-port="2" to-layer="7548" to-port="0" />
-		<edge from-layer="7457" from-port="2" to-layer="7515" to-port="0" />
-		<edge from-layer="7458" from-port="0" to-layer="7461" to-port="1" />
-		<edge from-layer="7459" from-port="0" to-layer="7461" to-port="2" />
-		<edge from-layer="7460" from-port="0" to-layer="7461" to-port="3" />
-		<edge from-layer="7461" from-port="4" to-layer="7476" to-port="0" />
-		<edge from-layer="7461" from-port="4" to-layer="7482" to-port="0" />
-		<edge from-layer="7461" from-port="4" to-layer="7465" to-port="0" />
-		<edge from-layer="7461" from-port="4" to-layer="7462" to-port="0" />
-		<edge from-layer="7462" from-port="2" to-layer="7485" to-port="0" />
-		<edge from-layer="7463" from-port="0" to-layer="7473" to-port="0" />
-		<edge from-layer="7464" from-port="0" to-layer="7473" to-port="1" />
-		<edge from-layer="7465" from-port="1" to-layer="7467" to-port="0" />
-		<edge from-layer="7466" from-port="0" to-layer="7467" to-port="2" />
-		<edge from-layer="7467" from-port="3" to-layer="7468" to-port="0" />
-		<edge from-layer="7468" from-port="2" to-layer="7469" to-port="0" />
-		<edge from-layer="7469" from-port="1" to-layer="7471" to-port="0" />
-		<edge from-layer="7470" from-port="0" to-layer="7471" to-port="1" />
-		<edge from-layer="7471" from-port="2" to-layer="7473" to-port="2" />
-		<edge from-layer="7472" from-port="0" to-layer="7473" to-port="3" />
-		<edge from-layer="7473" from-port="4" to-layer="7476" to-port="1" />
-		<edge from-layer="7473" from-port="4" to-layer="7482" to-port="2" />
-		<edge from-layer="7474" from-port="0" to-layer="7476" to-port="2" />
-		<edge from-layer="7475" from-port="0" to-layer="7476" to-port="3" />
-		<edge from-layer="7476" from-port="4" to-layer="7479" to-port="0" />
-		<edge from-layer="7477" from-port="0" to-layer="7478" to-port="0" />
-		<edge from-layer="7478" from-port="1" to-layer="7479" to-port="1" />
-		<edge from-layer="7479" from-port="2" to-layer="7483" to-port="0" />
-		<edge from-layer="7480" from-port="0" to-layer="7482" to-port="1" />
-		<edge from-layer="7481" from-port="0" to-layer="7482" to-port="3" />
-		<edge from-layer="7482" from-port="4" to-layer="7483" to-port="1" />
-		<edge from-layer="7483" from-port="2" to-layer="7484" to-port="0" />
-		<edge from-layer="7484" from-port="2" to-layer="7485" to-port="1" />
-		<edge from-layer="7485" from-port="2" to-layer="7490" to-port="0" />
-		<edge from-layer="7486" from-port="0" to-layer="7489" to-port="1" />
-		<edge from-layer="7487" from-port="0" to-layer="7489" to-port="2" />
-		<edge from-layer="7488" from-port="0" to-layer="7489" to-port="3" />
-		<edge from-layer="7489" from-port="4" to-layer="7490" to-port="1" />
-		<edge from-layer="7490" from-port="2" to-layer="7491" to-port="1" />
-		<edge from-layer="7491" from-port="2" to-layer="7512" to-port="0" />
-		<edge from-layer="7491" from-port="2" to-layer="7548" to-port="1" />
-		<edge from-layer="7491" from-port="2" to-layer="7674" to-port="0" />
-		<edge from-layer="7492" from-port="0" to-layer="7493" to-port="0" />
-		<edge from-layer="7493" from-port="1" to-layer="7498" to-port="0" />
-		<edge from-layer="7494" from-port="0" to-layer="7497" to-port="1" />
-		<edge from-layer="7495" from-port="0" to-layer="7497" to-port="2" />
-		<edge from-layer="7496" from-port="0" to-layer="7497" to-port="3" />
-		<edge from-layer="7497" from-port="4" to-layer="7498" to-port="1" />
-		<edge from-layer="7498" from-port="2" to-layer="7499" to-port="0" />
-		<edge from-layer="7499" from-port="1" to-layer="7501" to-port="0" />
-		<edge from-layer="7500" from-port="0" to-layer="7501" to-port="2" />
-		<edge from-layer="7501" from-port="3" to-layer="7508" to-port="0" />
-		<edge from-layer="7502" from-port="0" to-layer="7505" to-port="1" />
-		<edge from-layer="7503" from-port="0" to-layer="7505" to-port="2" />
-		<edge from-layer="7504" from-port="0" to-layer="7505" to-port="3" />
-		<edge from-layer="7505" from-port="4" to-layer="7507" to-port="0" />
-		<edge from-layer="7506" from-port="0" to-layer="7507" to-port="1" />
-		<edge from-layer="7507" from-port="2" to-layer="7508" to-port="1" />
-		<edge from-layer="7508" from-port="2" to-layer="7548" to-port="2" />
-		<edge from-layer="7508" from-port="2" to-layer="7673" to-port="0" />
-		<edge from-layer="7509" from-port="0" to-layer="7527" to-port="0" />
-		<edge from-layer="7510" from-port="0" to-layer="7522" to-port="0" />
-		<edge from-layer="7511" from-port="0" to-layer="7525" to-port="1" />
-		<edge from-layer="7511" from-port="0" to-layer="7522" to-port="1" />
-		<edge from-layer="7512" from-port="1" to-layer="7514" to-port="0" />
-		<edge from-layer="7513" from-port="0" to-layer="7514" to-port="2" />
-		<edge from-layer="7514" from-port="3" to-layer="7518" to-port="0" />
-		<edge from-layer="7514" from-port="3" to-layer="7524" to-port="0" />
-		<edge from-layer="7515" from-port="1" to-layer="7517" to-port="0" />
-		<edge from-layer="7515" from-port="1" to-layer="7543" to-port="0" />
-		<edge from-layer="7516" from-port="0" to-layer="7517" to-port="2" />
-		<edge from-layer="7517" from-port="3" to-layer="7518" to-port="1" />
-		<edge from-layer="7518" from-port="2" to-layer="7520" to-port="0" />
-		<edge from-layer="7519" from-port="0" to-layer="7520" to-port="1" />
-		<edge from-layer="7519" from-port="0" to-layer="7524" to-port="1" />
-		<edge from-layer="7520" from-port="2" to-layer="7522" to-port="2" />
-		<edge from-layer="7521" from-port="0" to-layer="7522" to-port="3" />
-		<edge from-layer="7521" from-port="0" to-layer="7525" to-port="3" />
-		<edge from-layer="7522" from-port="4" to-layer="7527" to-port="1" />
-		<edge from-layer="7523" from-port="0" to-layer="7525" to-port="0" />
-		<edge from-layer="7524" from-port="2" to-layer="7525" to-port="2" />
-		<edge from-layer="7524" from-port="2" to-layer="7532" to-port="2" />
-		<edge from-layer="7525" from-port="4" to-layer="7527" to-port="2" />
-		<edge from-layer="7526" from-port="0" to-layer="7527" to-port="3" />
-		<edge from-layer="7527" from-port="4" to-layer="7534" to-port="0" />
-		<edge from-layer="7528" from-port="0" to-layer="7534" to-port="1" />
-		<edge from-layer="7529" from-port="0" to-layer="7532" to-port="0" />
-		<edge from-layer="7530" from-port="0" to-layer="7532" to-port="1" />
-		<edge from-layer="7531" from-port="0" to-layer="7532" to-port="3" />
-		<edge from-layer="7532" from-port="4" to-layer="7534" to-port="2" />
-		<edge from-layer="7533" from-port="0" to-layer="7534" to-port="3" />
-		<edge from-layer="7534" from-port="4" to-layer="7538" to-port="0" />
-		<edge from-layer="7535" from-port="0" to-layer="7536" to-port="0" />
-		<edge from-layer="7536" from-port="1" to-layer="7538" to-port="1" />
-		<edge from-layer="7537" from-port="0" to-layer="7538" to-port="2" />
-		<edge from-layer="7538" from-port="3" to-layer="7539" to-port="0" />
-		<edge from-layer="7539" from-port="2" to-layer="7547" to-port="0" />
-		<edge from-layer="7539" from-port="2" to-layer="7540" to-port="0" />
-		<edge from-layer="7540" from-port="1" to-layer="7546" to-port="0" />
-		<edge from-layer="7541" from-port="0" to-layer="7543" to-port="1" />
-		<edge from-layer="7542" from-port="0" to-layer="7543" to-port="2" />
-		<edge from-layer="7543" from-port="3" to-layer="7544" to-port="0" />
-		<edge from-layer="7544" from-port="4" to-layer="7545" to-port="0" />
-		<edge from-layer="7545" from-port="1" to-layer="7546" to-port="1" />
-		<edge from-layer="7546" from-port="2" to-layer="7547" to-port="1" />
-		<edge from-layer="7547" from-port="2" to-layer="7548" to-port="3" />
-		<edge from-layer="7548" from-port="4" to-layer="7550" to-port="0" />
-		<edge from-layer="7549" from-port="0" to-layer="7550" to-port="1" />
-		<edge from-layer="7550" from-port="2" to-layer="7552" to-port="0" />
-		<edge from-layer="7551" from-port="0" to-layer="7552" to-port="1" />
-		<edge from-layer="7552" from-port="2" to-layer="7561" to-port="0" />
-		<edge from-layer="7553" from-port="0" to-layer="7554" to-port="0" />
-		<edge from-layer="7554" from-port="1" to-layer="7557" to-port="0" />
-		<edge from-layer="7555" from-port="0" to-layer="7556" to-port="0" />
-		<edge from-layer="7556" from-port="1" to-layer="7557" to-port="1" />
-		<edge from-layer="7557" from-port="2" to-layer="7559" to-port="0" />
-		<edge from-layer="7558" from-port="0" to-layer="7559" to-port="1" />
-		<edge from-layer="7559" from-port="2" to-layer="7560" to-port="0" />
-		<edge from-layer="7560" from-port="1" to-layer="7561" to-port="1" />
-		<edge from-layer="7561" from-port="2" to-layer="7564" to-port="0" />
-		<edge from-layer="7562" from-port="0" to-layer="7563" to-port="0" />
-		<edge from-layer="7563" from-port="1" to-layer="7564" to-port="1" />
-		<edge from-layer="7564" from-port="2" to-layer="7565" to-port="1" />
-		<edge from-layer="7565" from-port="2" to-layer="7566" to-port="0" />
-		<edge from-layer="7566" from-port="2" to-layer="7850" to-port="1" />
-		<edge from-layer="7566" from-port="2" to-layer="7568" to-port="0" />
-		<edge from-layer="7567" from-port="0" to-layer="7568" to-port="1" />
-		<edge from-layer="7568" from-port="2" to-layer="7725" to-port="0" />
-		<edge from-layer="7568" from-port="2" to-layer="7571" to-port="0" />
-		<edge from-layer="7569" from-port="0" to-layer="7570" to-port="0" />
-		<edge from-layer="7570" from-port="1" to-layer="7571" to-port="1" />
-		<edge from-layer="7571" from-port="2" to-layer="7574" to-port="0" />
-		<edge from-layer="7572" from-port="0" to-layer="7573" to-port="0" />
-		<edge from-layer="7573" from-port="1" to-layer="7574" to-port="1" />
-		<edge from-layer="7574" from-port="2" to-layer="7583" to-port="0" />
-		<edge from-layer="7575" from-port="0" to-layer="7576" to-port="0" />
-		<edge from-layer="7576" from-port="1" to-layer="7579" to-port="0" />
-		<edge from-layer="7577" from-port="0" to-layer="7578" to-port="0" />
-		<edge from-layer="7578" from-port="1" to-layer="7579" to-port="1" />
-		<edge from-layer="7579" from-port="2" to-layer="7581" to-port="0" />
-		<edge from-layer="7580" from-port="0" to-layer="7581" to-port="1" />
-		<edge from-layer="7581" from-port="2" to-layer="7582" to-port="0" />
-		<edge from-layer="7582" from-port="1" to-layer="7583" to-port="1" />
-		<edge from-layer="7583" from-port="2" to-layer="7586" to-port="0" />
-		<edge from-layer="7584" from-port="0" to-layer="7585" to-port="0" />
-		<edge from-layer="7585" from-port="1" to-layer="7586" to-port="1" />
-		<edge from-layer="7586" from-port="2" to-layer="7588" to-port="0" />
-		<edge from-layer="7587" from-port="0" to-layer="7588" to-port="1" />
-		<edge from-layer="7588" from-port="2" to-layer="7592" to-port="0" />
-		<edge from-layer="7588" from-port="2" to-layer="7610" to-port="0" />
-		<edge from-layer="7588" from-port="2" to-layer="7757" to-port="0" />
-		<edge from-layer="7589" from-port="0" to-layer="7592" to-port="1" />
-		<edge from-layer="7590" from-port="0" to-layer="7592" to-port="2" />
-		<edge from-layer="7591" from-port="0" to-layer="7592" to-port="3" />
-		<edge from-layer="7592" from-port="4" to-layer="7594" to-port="0" />
-		<edge from-layer="7593" from-port="0" to-layer="7594" to-port="1" />
-		<edge from-layer="7594" from-port="2" to-layer="7595" to-port="1" />
-		<edge from-layer="7595" from-port="2" to-layer="7832" to-port="2" />
-		<edge from-layer="7595" from-port="2" to-layer="7596" to-port="0" />
-		<edge from-layer="7597" from-port="0" to-layer="7598" to-port="0" />
-		<edge from-layer="7598" from-port="1" to-layer="7603" to-port="0" />
-		<edge from-layer="7599" from-port="0" to-layer="7602" to-port="1" />
-		<edge from-layer="7600" from-port="0" to-layer="7602" to-port="2" />
-		<edge from-layer="7601" from-port="0" to-layer="7602" to-port="3" />
-		<edge from-layer="7602" from-port="4" to-layer="7603" to-port="1" />
-		<edge from-layer="7603" from-port="2" to-layer="7604" to-port="0" />
-		<edge from-layer="7604" from-port="1" to-layer="7606" to-port="0" />
-		<edge from-layer="7605" from-port="0" to-layer="7606" to-port="2" />
-		<edge from-layer="7606" from-port="3" to-layer="7623" to-port="0" />
-		<edge from-layer="7606" from-port="3" to-layer="7671" to-port="0" />
-		<edge from-layer="7607" from-port="0" to-layer="7610" to-port="1" />
-		<edge from-layer="7608" from-port="0" to-layer="7610" to-port="2" />
-		<edge from-layer="7609" from-port="0" to-layer="7610" to-port="3" />
-		<edge from-layer="7610" from-port="4" to-layer="7612" to-port="0" />
-		<edge from-layer="7611" from-port="0" to-layer="7612" to-port="1" />
-		<edge from-layer="7612" from-port="2" to-layer="7616" to-port="0" />
-		<edge from-layer="7612" from-port="2" to-layer="7669" to-port="0" />
-		<edge from-layer="7612" from-port="2" to-layer="7620" to-port="0" />
-		<edge from-layer="7613" from-port="0" to-layer="7616" to-port="1" />
-		<edge from-layer="7614" from-port="0" to-layer="7616" to-port="2" />
-		<edge from-layer="7615" from-port="0" to-layer="7616" to-port="3" />
-		<edge from-layer="7616" from-port="4" to-layer="7648" to-port="0" />
-		<edge from-layer="7616" from-port="4" to-layer="7637" to-port="0" />
-		<edge from-layer="7616" from-port="4" to-layer="7634" to-port="0" />
-		<edge from-layer="7616" from-port="4" to-layer="7654" to-port="0" />
-		<edge from-layer="7617" from-port="0" to-layer="7618" to-port="0" />
-		<edge from-layer="7618" from-port="1" to-layer="7630" to-port="0" />
-		<edge from-layer="7619" from-port="0" to-layer="7630" to-port="1" />
-		<edge from-layer="7620" from-port="1" to-layer="7622" to-port="0" />
-		<edge from-layer="7621" from-port="0" to-layer="7622" to-port="2" />
-		<edge from-layer="7622" from-port="3" to-layer="7626" to-port="0" />
-		<edge from-layer="7623" from-port="1" to-layer="7625" to-port="0" />
-		<edge from-layer="7624" from-port="0" to-layer="7625" to-port="2" />
-		<edge from-layer="7625" from-port="3" to-layer="7626" to-port="1" />
-		<edge from-layer="7626" from-port="2" to-layer="7628" to-port="0" />
-		<edge from-layer="7627" from-port="0" to-layer="7628" to-port="1" />
-		<edge from-layer="7628" from-port="2" to-layer="7630" to-port="2" />
-		<edge from-layer="7628" from-port="2" to-layer="7660" to-port="2" />
-		<edge from-layer="7629" from-port="0" to-layer="7630" to-port="3" />
-		<edge from-layer="7630" from-port="4" to-layer="7632" to-port="0" />
-		<edge from-layer="7631" from-port="0" to-layer="7632" to-port="2" />
-		<edge from-layer="7632" from-port="3" to-layer="7633" to-port="0" />
-		<edge from-layer="7633" from-port="2" to-layer="7634" to-port="1" />
-		<edge from-layer="7633" from-port="2" to-layer="7764" to-port="1" />
-		<edge from-layer="7634" from-port="2" to-layer="7665" to-port="0" />
-		<edge from-layer="7635" from-port="0" to-layer="7645" to-port="0" />
-		<edge from-layer="7636" from-port="0" to-layer="7645" to-port="1" />
-		<edge from-layer="7637" from-port="1" to-layer="7639" to-port="0" />
-		<edge from-layer="7638" from-port="0" to-layer="7639" to-port="2" />
-		<edge from-layer="7639" from-port="3" to-layer="7640" to-port="0" />
-		<edge from-layer="7640" from-port="2" to-layer="7641" to-port="0" />
-		<edge from-layer="7641" from-port="1" to-layer="7643" to-port="0" />
-		<edge from-layer="7642" from-port="0" to-layer="7643" to-port="1" />
-		<edge from-layer="7643" from-port="2" to-layer="7645" to-port="2" />
-		<edge from-layer="7644" from-port="0" to-layer="7645" to-port="3" />
-		<edge from-layer="7645" from-port="4" to-layer="7648" to-port="1" />
-		<edge from-layer="7645" from-port="4" to-layer="7654" to-port="2" />
-		<edge from-layer="7646" from-port="0" to-layer="7648" to-port="2" />
-		<edge from-layer="7647" from-port="0" to-layer="7648" to-port="3" />
-		<edge from-layer="7648" from-port="4" to-layer="7651" to-port="0" />
-		<edge from-layer="7649" from-port="0" to-layer="7650" to-port="0" />
-		<edge from-layer="7650" from-port="1" to-layer="7651" to-port="1" />
-		<edge from-layer="7651" from-port="2" to-layer="7655" to-port="0" />
-		<edge from-layer="7652" from-port="0" to-layer="7654" to-port="1" />
-		<edge from-layer="7653" from-port="0" to-layer="7654" to-port="3" />
-		<edge from-layer="7654" from-port="4" to-layer="7655" to-port="1" />
-		<edge from-layer="7655" from-port="2" to-layer="7664" to-port="0" />
-		<edge from-layer="7656" from-port="0" to-layer="7657" to-port="0" />
-		<edge from-layer="7657" from-port="1" to-layer="7660" to-port="0" />
-		<edge from-layer="7658" from-port="0" to-layer="7660" to-port="1" />
-		<edge from-layer="7659" from-port="0" to-layer="7660" to-port="3" />
-		<edge from-layer="7660" from-port="4" to-layer="7662" to-port="0" />
-		<edge from-layer="7661" from-port="0" to-layer="7662" to-port="2" />
-		<edge from-layer="7662" from-port="3" to-layer="7663" to-port="0" />
-		<edge from-layer="7663" from-port="2" to-layer="7786" to-port="1" />
-		<edge from-layer="7663" from-port="2" to-layer="7664" to-port="1" />
-		<edge from-layer="7664" from-port="2" to-layer="7665" to-port="1" />
-		<edge from-layer="7665" from-port="2" to-layer="7670" to-port="0" />
-		<edge from-layer="7666" from-port="0" to-layer="7669" to-port="1" />
-		<edge from-layer="7667" from-port="0" to-layer="7669" to-port="2" />
-		<edge from-layer="7668" from-port="0" to-layer="7669" to-port="3" />
-		<edge from-layer="7669" from-port="4" to-layer="7670" to-port="1" />
-		<edge from-layer="7670" from-port="2" to-layer="7671" to-port="1" />
-		<edge from-layer="7671" from-port="2" to-layer="7796" to-port="0" />
-		<edge from-layer="7671" from-port="2" to-layer="7672" to-port="0" />
-		<edge from-layer="7671" from-port="2" to-layer="7832" to-port="1" />
-		<edge from-layer="7723" from-port="0" to-layer="7724" to-port="0" />
-		<edge from-layer="7724" from-port="1" to-layer="7725" to-port="1" />
-		<edge from-layer="7725" from-port="2" to-layer="7728" to-port="0" />
-		<edge from-layer="7726" from-port="0" to-layer="7727" to-port="0" />
-		<edge from-layer="7727" from-port="1" to-layer="7728" to-port="1" />
-		<edge from-layer="7728" from-port="2" to-layer="7737" to-port="0" />
-		<edge from-layer="7729" from-port="0" to-layer="7730" to-port="0" />
-		<edge from-layer="7730" from-port="1" to-layer="7733" to-port="0" />
-		<edge from-layer="7731" from-port="0" to-layer="7732" to-port="0" />
-		<edge from-layer="7732" from-port="1" to-layer="7733" to-port="1" />
-		<edge from-layer="7733" from-port="2" to-layer="7735" to-port="0" />
-		<edge from-layer="7734" from-port="0" to-layer="7735" to-port="1" />
-		<edge from-layer="7735" from-port="2" to-layer="7736" to-port="0" />
-		<edge from-layer="7736" from-port="1" to-layer="7737" to-port="1" />
-		<edge from-layer="7737" from-port="2" to-layer="7740" to-port="0" />
-		<edge from-layer="7738" from-port="0" to-layer="7739" to-port="0" />
-		<edge from-layer="7739" from-port="1" to-layer="7740" to-port="1" />
-		<edge from-layer="7740" from-port="2" to-layer="7741" to-port="0" />
-		<edge from-layer="7741" from-port="1" to-layer="7750" to-port="0" />
-		<edge from-layer="7742" from-port="0" to-layer="7743" to-port="0" />
-		<edge from-layer="7743" from-port="1" to-layer="7746" to-port="0" />
-		<edge from-layer="7744" from-port="0" to-layer="7745" to-port="0" />
-		<edge from-layer="7745" from-port="1" to-layer="7746" to-port="1" />
-		<edge from-layer="7746" from-port="2" to-layer="7748" to-port="0" />
-		<edge from-layer="7747" from-port="0" to-layer="7748" to-port="1" />
-		<edge from-layer="7748" from-port="2" to-layer="7749" to-port="0" />
-		<edge from-layer="7749" from-port="1" to-layer="7750" to-port="1" />
-		<edge from-layer="7750" from-port="2" to-layer="7753" to-port="0" />
-		<edge from-layer="7751" from-port="0" to-layer="7752" to-port="0" />
-		<edge from-layer="7752" from-port="1" to-layer="7753" to-port="1" />
-		<edge from-layer="7753" from-port="2" to-layer="7849" to-port="0" />
-		<edge from-layer="7754" from-port="0" to-layer="7757" to-port="1" />
-		<edge from-layer="7755" from-port="0" to-layer="7757" to-port="2" />
-		<edge from-layer="7756" from-port="0" to-layer="7757" to-port="3" />
-		<edge from-layer="7757" from-port="4" to-layer="7759" to-port="0" />
-		<edge from-layer="7758" from-port="0" to-layer="7759" to-port="1" />
-		<edge from-layer="7759" from-port="2" to-layer="7763" to-port="0" />
-		<edge from-layer="7759" from-port="2" to-layer="7791" to-port="0" />
-		<edge from-layer="7760" from-port="0" to-layer="7763" to-port="1" />
-		<edge from-layer="7761" from-port="0" to-layer="7763" to-port="2" />
-		<edge from-layer="7762" from-port="0" to-layer="7763" to-port="3" />
-		<edge from-layer="7763" from-port="4" to-layer="7778" to-port="0" />
-		<edge from-layer="7763" from-port="4" to-layer="7764" to-port="0" />
-		<edge from-layer="7763" from-port="4" to-layer="7767" to-port="0" />
-		<edge from-layer="7763" from-port="4" to-layer="7784" to-port="0" />
-		<edge from-layer="7764" from-port="2" to-layer="7787" to-port="0" />
-		<edge from-layer="7765" from-port="0" to-layer="7775" to-port="0" />
-		<edge from-layer="7766" from-port="0" to-layer="7775" to-port="1" />
-		<edge from-layer="7767" from-port="1" to-layer="7769" to-port="0" />
-		<edge from-layer="7768" from-port="0" to-layer="7769" to-port="2" />
-		<edge from-layer="7769" from-port="3" to-layer="7770" to-port="0" />
-		<edge from-layer="7770" from-port="2" to-layer="7771" to-port="0" />
-		<edge from-layer="7771" from-port="1" to-layer="7773" to-port="0" />
-		<edge from-layer="7772" from-port="0" to-layer="7773" to-port="1" />
-		<edge from-layer="7773" from-port="2" to-layer="7775" to-port="2" />
-		<edge from-layer="7774" from-port="0" to-layer="7775" to-port="3" />
-		<edge from-layer="7775" from-port="4" to-layer="7778" to-port="1" />
-		<edge from-layer="7775" from-port="4" to-layer="7784" to-port="2" />
-		<edge from-layer="7776" from-port="0" to-layer="7778" to-port="2" />
-		<edge from-layer="7777" from-port="0" to-layer="7778" to-port="3" />
-		<edge from-layer="7778" from-port="4" to-layer="7781" to-port="0" />
-		<edge from-layer="7779" from-port="0" to-layer="7780" to-port="0" />
-		<edge from-layer="7780" from-port="1" to-layer="7781" to-port="1" />
-		<edge from-layer="7781" from-port="2" to-layer="7785" to-port="0" />
-		<edge from-layer="7782" from-port="0" to-layer="7784" to-port="1" />
-		<edge from-layer="7783" from-port="0" to-layer="7784" to-port="3" />
-		<edge from-layer="7784" from-port="4" to-layer="7785" to-port="1" />
-		<edge from-layer="7785" from-port="2" to-layer="7786" to-port="0" />
-		<edge from-layer="7786" from-port="2" to-layer="7787" to-port="1" />
-		<edge from-layer="7787" from-port="2" to-layer="7792" to-port="0" />
-		<edge from-layer="7788" from-port="0" to-layer="7791" to-port="1" />
-		<edge from-layer="7789" from-port="0" to-layer="7791" to-port="2" />
-		<edge from-layer="7790" from-port="0" to-layer="7791" to-port="3" />
-		<edge from-layer="7791" from-port="4" to-layer="7792" to-port="1" />
-		<edge from-layer="7792" from-port="2" to-layer="7799" to-port="0" />
-		<edge from-layer="7792" from-port="2" to-layer="7832" to-port="0" />
-		<edge from-layer="7793" from-port="0" to-layer="7811" to-port="0" />
-		<edge from-layer="7794" from-port="0" to-layer="7806" to-port="0" />
-		<edge from-layer="7795" from-port="0" to-layer="7806" to-port="1" />
-		<edge from-layer="7795" from-port="0" to-layer="7809" to-port="1" />
-		<edge from-layer="7796" from-port="1" to-layer="7798" to-port="0" />
-		<edge from-layer="7797" from-port="0" to-layer="7798" to-port="2" />
-		<edge from-layer="7798" from-port="3" to-layer="7802" to-port="0" />
-		<edge from-layer="7798" from-port="3" to-layer="7808" to-port="0" />
-		<edge from-layer="7799" from-port="1" to-layer="7801" to-port="0" />
-		<edge from-layer="7799" from-port="1" to-layer="7827" to-port="0" />
-		<edge from-layer="7800" from-port="0" to-layer="7801" to-port="2" />
-		<edge from-layer="7801" from-port="3" to-layer="7802" to-port="1" />
-		<edge from-layer="7802" from-port="2" to-layer="7804" to-port="0" />
-		<edge from-layer="7803" from-port="0" to-layer="7804" to-port="1" />
-		<edge from-layer="7803" from-port="0" to-layer="7808" to-port="1" />
-		<edge from-layer="7804" from-port="2" to-layer="7806" to-port="2" />
-		<edge from-layer="7805" from-port="0" to-layer="7806" to-port="3" />
-		<edge from-layer="7805" from-port="0" to-layer="7809" to-port="3" />
-		<edge from-layer="7806" from-port="4" to-layer="7811" to-port="1" />
-		<edge from-layer="7807" from-port="0" to-layer="7809" to-port="0" />
-		<edge from-layer="7808" from-port="2" to-layer="7809" to-port="2" />
-		<edge from-layer="7808" from-port="2" to-layer="7816" to-port="2" />
-		<edge from-layer="7809" from-port="4" to-layer="7811" to-port="2" />
-		<edge from-layer="7810" from-port="0" to-layer="7811" to-port="3" />
-		<edge from-layer="7811" from-port="4" to-layer="7818" to-port="0" />
-		<edge from-layer="7812" from-port="0" to-layer="7818" to-port="1" />
-		<edge from-layer="7813" from-port="0" to-layer="7816" to-port="0" />
-		<edge from-layer="7814" from-port="0" to-layer="7816" to-port="1" />
-		<edge from-layer="7815" from-port="0" to-layer="7816" to-port="3" />
-		<edge from-layer="7816" from-port="4" to-layer="7818" to-port="2" />
-		<edge from-layer="7817" from-port="0" to-layer="7818" to-port="3" />
-		<edge from-layer="7818" from-port="4" to-layer="7822" to-port="0" />
-		<edge from-layer="7819" from-port="0" to-layer="7820" to-port="0" />
-		<edge from-layer="7820" from-port="1" to-layer="7822" to-port="1" />
-		<edge from-layer="7821" from-port="0" to-layer="7822" to-port="2" />
-		<edge from-layer="7822" from-port="3" to-layer="7823" to-port="0" />
-		<edge from-layer="7823" from-port="2" to-layer="7831" to-port="0" />
-		<edge from-layer="7823" from-port="2" to-layer="7824" to-port="0" />
-		<edge from-layer="7824" from-port="1" to-layer="7830" to-port="0" />
-		<edge from-layer="7825" from-port="0" to-layer="7827" to-port="1" />
-		<edge from-layer="7826" from-port="0" to-layer="7827" to-port="2" />
-		<edge from-layer="7827" from-port="3" to-layer="7828" to-port="0" />
-		<edge from-layer="7828" from-port="4" to-layer="7829" to-port="0" />
-		<edge from-layer="7829" from-port="1" to-layer="7830" to-port="1" />
-		<edge from-layer="7830" from-port="2" to-layer="7831" to-port="1" />
-		<edge from-layer="7831" from-port="2" to-layer="7832" to-port="3" />
-		<edge from-layer="7832" from-port="4" to-layer="7834" to-port="0" />
-		<edge from-layer="7833" from-port="0" to-layer="7834" to-port="1" />
-		<edge from-layer="7834" from-port="2" to-layer="7836" to-port="0" />
-		<edge from-layer="7835" from-port="0" to-layer="7836" to-port="1" />
-		<edge from-layer="7836" from-port="2" to-layer="7845" to-port="0" />
-		<edge from-layer="7837" from-port="0" to-layer="7838" to-port="0" />
-		<edge from-layer="7838" from-port="1" to-layer="7841" to-port="0" />
-		<edge from-layer="7839" from-port="0" to-layer="7840" to-port="0" />
-		<edge from-layer="7840" from-port="1" to-layer="7841" to-port="1" />
-		<edge from-layer="7841" from-port="2" to-layer="7843" to-port="0" />
-		<edge from-layer="7842" from-port="0" to-layer="7843" to-port="1" />
-		<edge from-layer="7843" from-port="2" to-layer="7844" to-port="0" />
-		<edge from-layer="7844" from-port="1" to-layer="7845" to-port="1" />
-		<edge from-layer="7845" from-port="2" to-layer="7848" to-port="0" />
-		<edge from-layer="7846" from-port="0" to-layer="7847" to-port="0" />
-		<edge from-layer="7847" from-port="1" to-layer="7848" to-port="1" />
-		<edge from-layer="7848" from-port="2" to-layer="7849" to-port="1" />
-		<edge from-layer="7849" from-port="2" to-layer="7850" to-port="0" />
-		<edge from-layer="7850" from-port="2" to-layer="7852" to-port="0" />
-		<edge from-layer="7851" from-port="0" to-layer="7852" to-port="1" />
-		<edge from-layer="7852" from-port="2" to-layer="7855" to-port="0" />
-		<edge from-layer="7853" from-port="0" to-layer="7854" to-port="0" />
-		<edge from-layer="7854" from-port="1" to-layer="7855" to-port="1" />
-		<edge from-layer="7855" from-port="2" to-layer="7858" to-port="0" />
-		<edge from-layer="7856" from-port="0" to-layer="7857" to-port="0" />
-		<edge from-layer="7857" from-port="1" to-layer="7858" to-port="1" />
-		<edge from-layer="7858" from-port="2" to-layer="7867" to-port="0" />
-		<edge from-layer="7859" from-port="0" to-layer="7860" to-port="0" />
-		<edge from-layer="7860" from-port="1" to-layer="7863" to-port="0" />
-		<edge from-layer="7861" from-port="0" to-layer="7862" to-port="0" />
-		<edge from-layer="7862" from-port="1" to-layer="7863" to-port="1" />
-		<edge from-layer="7863" from-port="2" to-layer="7865" to-port="0" />
-		<edge from-layer="7864" from-port="0" to-layer="7865" to-port="1" />
-		<edge from-layer="7865" from-port="2" to-layer="7866" to-port="0" />
-		<edge from-layer="7866" from-port="1" to-layer="7867" to-port="1" />
-		<edge from-layer="7867" from-port="2" to-layer="7868" to-port="0" />
+		<edge from-layer="5237" from-port="2" to-layer="5238" to-port="0" />
+		<edge from-layer="5238" from-port="1" to-layer="5239" to-port="1" />
+		<edge from-layer="5239" from-port="2" to-layer="5241" to-port="0" />
+		<edge from-layer="5240" from-port="0" to-layer="5241" to-port="1" />
+		<edge from-layer="5241" from-port="2" to-layer="5242" to-port="1" />
+		<edge from-layer="5242" from-port="2" to-layer="5243" to-port="0" />
+		<edge from-layer="5243" from-port="2" to-layer="5245" to-port="0" />
+		<edge from-layer="5244" from-port="0" to-layer="5245" to-port="1" />
+		<edge from-layer="5245" from-port="2" to-layer="5247" to-port="0" />
+		<edge from-layer="5246" from-port="0" to-layer="5247" to-port="1" />
+		<edge from-layer="5247" from-port="2" to-layer="5249" to-port="0" />
+		<edge from-layer="5248" from-port="0" to-layer="5249" to-port="1" />
+		<edge from-layer="5249" from-port="2" to-layer="5258" to-port="0" />
+		<edge from-layer="5250" from-port="0" to-layer="5251" to-port="0" />
+		<edge from-layer="5251" from-port="1" to-layer="5254" to-port="0" />
+		<edge from-layer="5252" from-port="0" to-layer="5253" to-port="0" />
+		<edge from-layer="5253" from-port="1" to-layer="5254" to-port="1" />
+		<edge from-layer="5254" from-port="2" to-layer="5256" to-port="0" />
+		<edge from-layer="5255" from-port="0" to-layer="5256" to-port="1" />
+		<edge from-layer="5256" from-port="2" to-layer="5257" to-port="0" />
+		<edge from-layer="5257" from-port="1" to-layer="5258" to-port="1" />
+		<edge from-layer="5258" from-port="2" to-layer="5259" to-port="0" />
 	</edges>
 	<rt_info>
-		<Runtime_version value="2024.2.0-15229-1b5a6c49e9b" />
+		<Runtime_version value="2024.5.0-16901-32aaa2fbd96" />
 		<conversion_parameters>
 			<framework value="pytorch" />
 			<is_python_object value="True" />
@@ -130312,26 +86829,24 @@
 		<nncf>
 			<friendly_names_were_updated value="True" />
 			<weight_compression>
+				<advanced_parameters value="{'awq_params': {'subset_size': 32, 'percent_to_apply': 0.002, 'alpha_min': 0.0, 'alpha_max': 1.0, 'steps': 100}, 'scale_estimation_params': {'subset_size': 64, 'initial_steps': 5, 'scale_steps': 5, 'weight_penalty': -1.0}, 'gptq_params': {'damp_percent': 0.1, 'block_size': 128, 'subset_size': 128}, 'lora_correction_params': {'adapter_rank': 8, 'num_iterations': 3, 'apply_regularization': True, 'subset_size': 128, 'use_int8_adapters': True}}" />
 				<all_layers value="False" />
 				<awq value="False" />
+				<gptq value="False" />
 				<group_size value="-1" />
-				<ignored_scope>
-					<names value="[]" />
-					<patterns value="[]" />
-					<subgraphs value="[]" />
-					<types value="[]" />
-					<validate value="True" />
-				</ignored_scope>
+				<ignored_scope value="[]" />
+				<lora_correction value="False" />
 				<mode value="int8_asym" />
 				<ratio value="1" />
+				<scale_estimation value="False" />
 				<sensitivity_metric value="weight_quantization_error" />
 			</weight_compression>
 		</nncf>
 		<optimum>
-			<optimum_intel_version value="1.17.0.dev0+e1b6a59" />
-			<optimum_version value="1.19.1" />
-			<pytorch_version value="2.3.0+cpu" />
-			<transformers_version value="4.40.1" />
+			<optimum_intel_version value="1.20.0.dev0+2559620" />
+			<optimum_version value="1.23.1" />
+			<pytorch_version value="2.4.1+cpu" />
+			<transformers_version value="4.45.2" />
 		</optimum>
 	</rt_info>
 </net>